EA047516B1 - METHODS AND SYSTEMS FOR ANALYZING CHROMATOGRAPHIC DATA - Google Patents

Description

Cross-reference to related applications

This application claims priority to U.S. Patent Application No. 62/412,563, filed October 25, 2016, the entire disclosure of which is incorporated herein by reference.

Field of technology to which the invention relates

Aspects of the present invention relate generally to chromatography systems and methods and, in particular, to embodiments of methods and systems for analyzing chromatographic data, such as for in-process monitoring and control of chromatographic systems.

State of the art

Packed bed chromatographic processes play an important role in the production of biological medicinal products. Many active biologicals, such as proteins, are purified for use in pharmaceutical formulations using packed bed chromatography. Consequently, the performance of the chromatography column can have a significant impact on the critical process parameters (CPPs) and critical quality attributes (CQAs). In addition, the complexity and size of biologicals compared to, for example, small molecules, can significantly complicate the analysis of the quality and purity of the biological. Therefore, monitoring the quality, consistency, and integrity of chromatography processes and equipment using in-process control is important to ensure that product quality meets any applicable standards (e.g., government regulations).

Typically, column integrity can be determined by a uniform flow of plugged mobile phase through the stationary phase of the column (e.g., resin). Examples of loss of column integrity may include, for example, evidence of channeling, headspace, fouled flow areas, and the like. Channeling may occur when, among other things, the mobile phase is allowed to travel some distance from the column inlet to the column outlet without contacting the stationary phase. Headspace may refer to, among other things, when a lateral zone is created in the column that allows the mobile phase to flow without plugging. Fouled flow areas may include dirt or other residues on the surfaces of the inlet frit or outlet frit or in the pores of the resin.

There are several techniques for monitoring the performance and integrity of a chromatographic column. Some methods, such as the pulse injection method used to measure the height equivalent to a theoretical plate (HETP), require buffer solutions that must be specially prepared. Pulse injection methods typically require the chromatographic equipment and column to be operated outside of normal processes, which increases process time and labor costs. Other methods involve monitoring critical parameters (e.g., step yield, pre-pool volume, and maximum absorbance during loading) as part of normal production. However, setting alarm limits based on these parameters is complex and imprecise, and may result in false alarms or limits that are too wide.

There is a need for methods, systems and processes to measure and control column performance and integrity with accuracy and precision and with minimal process disruption. In addition, due to the inherent differences between chromatographic columns, chromatographic column runs and/or production lots of any given product being chromatographically analyzed, there is a need for methods, systems and processes by which the column performance and integrity analysis can be tailored to a specific column or columns, a specific run or runs and/or a specific lot or lots of product. Finally, there is a need for accurate in-process controls that utilize such analysis and for methods and systems to respond to deviations from such controls so that column integrity and performance problems can be identified and corrected early with minimal disruption and cost.

The essence of the invention

Embodiments of the present invention may be directed to a method for controlling a process that includes: obtaining raw chromatographic data that includes a plurality of signals, where each signal from the plurality of signals is associated with one of a plurality of blocks; obtaining a subset of data by selecting a combination of a first block and a first signal from the raw chromatographic data; generating processed chromatographic data by applying a noise reduction method to the subset of data; generating transition data by performing transition analysis on the processed chromatographic data; and performing an action based on the transition data.

In some embodiments, the method may further comprise performing an analysis on a chromatographic column, where raw chromatographic data may be obtained from the analysis on the chromatographic column. In still other embodiments, the raw chromatographic data may be obtained from a chromatographic process unit. In still other embodiments, each block of the plurality of blocks may correspond to a step in the chromatographic process. In further embodiments, the selected combination may comprise a first block, a first signal, and a second signal of the plurality of signals.

In still other embodiments, the method may also include selecting a combination of the first block and the first signal according to a profile defining a plurality of selection criteria. In some embodiments, the plurality of selection criteria may include: the extent to which the blocks appear at regular intervals in the chromatographic cycle; the extent to which one of the plurality of signals saturates the detector; the extent to which the plurality of signals approaches a stationary phase at a particular level; the amount of variation of the plurality of signals; and/or the number of inflection points exhibited by the plurality of signals during the transition phase.

In some embodiments, selecting a combination of the first block and the first signal may include selecting a primary combination of the block and signal, and the method may further include selecting a secondary combination of the block and signal. In further embodiments, the noise reduction method may include: selecting a portion of a subset of data for analysis using predetermined control points; normalizing the portion to prevent the magnitude of the deviation; using at least one smoothing filter on the portion to generate smoothed data; and analyzing the portion for a dynamic signal error. In yet other embodiments, the method may further include: selecting smoothed data corresponding to a feature of a chromatographic transition, where the feature includes one of: a duration of the derivative; a maximum intensity; a duration from initiation; or an expected background noise of the sensor. In yet other embodiments, analyzing the transition may include generating a curve using the processed chromatographic data and analyzing the curve to generate performance parameters.

In some embodiments, the method may further comprise generating an individual value control chart, a moving range control chart, or a range control chart based on the transition data and generating performance data by applying statistical process control to the individual value control chart, the moving range control chart, or the range control chart, where performing an action based on the transition data may comprise performing an action based on the performance data. In some embodiments, applying statistical process control may comprise performing one of multivariate data analysis or principal component analysis. In some embodiments, performing an action based on the performance data may comprise generating an event notification, generating an event assessment, or generating a rejection notification form. Some embodiments of the present disclosure may comprise a chromatography method that comprises performing a process control method during analysis on a chromatography column.

Some aspects of the present invention may relate to a method for controlling a process, including: obtaining a sample of raw chromatographic data; generating smoothed data by applying a noise reduction method to the sample of raw chromatographic data, generating processed chromatographic data by selecting smoothed data corresponding to a chromatographic transition feature, and performing an action based on the processed chromatographic data. The noise reduction technique may include selecting a portion of the smoothed data for analysis using predetermined values, normalizing the portion of the data to prevent the amount of deviation, using at least one smoothing filter on the portion of the data to generate the smoothed data, and analyzing the portion of the data for dynamic signal errors.

In some embodiments, obtaining a sample of raw chromatographic data may include obtaining raw chromatographic data including a plurality of signals and a plurality of blocks, where each signal of the plurality of signals may be associated with a block, and selecting a combination of a first block and a first signal from the raw chromatographic data.

In some embodiments, the method may further include using the processed chromatographic data to generate one of a control chart of individual values, a control chart of moving ranges, or a control chart of ranges, and generating performance data by applying statistical process control to the control chart of individual values, the control chart of moving ranges, or the control chart of ranges by performing a multivariate data analysis or a principal component analysis. In some embodiments, performing an action based on the processed chromatographic data may include performing an action based on the performance data. In some embodiments, the action may include generating an event notification, generating an event assessment, or generating a rejection notification form.

Some aspects of the present invention may include a method of monitoring a process, wherein the method includes obtaining processed chromatographic data including a combination of a first block and a first signal, performing a transition analysis on the processed chromatographic data, generating one of an individual value control chart-a moving range control chart-a range control chart (I-MR-R) based on the transition analysis,

- 2 047516 Analyzing the performance data using the multivariate data analysis method for the I-MR-R chart and performing actions based on the performance data. The action may include one of generating an event notification, generating an event assessment, or generating a rejection notification form.

In some embodiments, the processed chromatographic data may include a sample of raw chromatographic data to which a noise reduction method has been applied. In some embodiments, the sample of raw chromatographic data may be obtained from a chromatographic process unit.

Brief description of the figures

The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate the disclosed embodiments and, together with the description, serve to explain the principles of the disclosed embodiments.

Fig. 1 is a schematic representation of an exemplary chromatographic system with which various embodiments of the present invention may be carried out.

Fig. 2 shows an approximate chromatogram.

Fig. 3 shows an approximate normalized graph of the upward chromatographic transition.

Fig. 4 is a graph of the increasing chromatographic transitions of the conductivity equilibration units for three batches according to certain aspects of the present invention.

Fig. 5 illustrates an exemplary process for analyzing chromatographic data and performing process control according to certain aspects of the present invention.

Fig. 6 illustrates another exemplary process for analyzing chromatographic data and performing process control in accordance with certain aspects of the present invention.

Fig. 7 illustrates an exemplary data file in accordance with certain aspects of the present invention.

Fig. 8 is an exemplary loading schedule for a multifactor model in accordance with certain aspects of the present invention.

Fig. 9 illustrates an exemplary data smoothing process according to certain aspects of the present invention.

Fig. 10 is a graph of the loading of each variable in a principal component from 27 batches according to certain aspects of the present invention.

Fig. 11 is an exemplary plot of scores from 27 games in accordance with certain aspects of the present invention.

Fig. 12 is an exemplary loading schedule for a multifactor model in accordance with certain aspects of the present invention.

Fig. 13 is an exemplary score chart according to certain aspects of the present invention.

Fig. 14 is a control chart of individual values for the asymmetry coefficient during operation of a particular chromatographic apparatus according to certain aspects of the present invention.

Fig. 15 is a control chart of the sliding ranges for the asymmetry coefficient during operation of a particular chromatographic apparatus according to certain aspects of the present invention.

Fig. 16 is a control chart of the range of the asymmetry coefficient during operation of a particular chromatographic apparatus according to certain aspects of the present invention.

Fig. 17 is a control chart of individual values for a non-Gaussian HETP distribution (NG-HETP) according to certain aspects of the present invention.

Fig. 18 is a control chart of sliding swings for NG-HETP according to certain aspects of the present invention.

Fig. 19 is a span control chart for NG-HETP according to certain aspects of the present invention.

Fig. 20 is another control chart of individual values for NG-HETP according to certain aspects of the present invention.

Fig. 21 shows another control chart of individual values for NG-HETP according to certain aspects of the present invention.

Fig. 22 illustrates an exemplary system with which aspects of the present disclosure may be implemented.

Fig. 23 illustrates an exemplary user interface according to certain aspects of the present invention.

Fig. 24 shows an exemplary report according to some aspects of the present invention.

- 3 047516

Detailed description of the invention

The present invention relates to improvements in drug product manufacturing and laboratory methods, as well as improvements in computer functionality associated with drug product manufacturing and laboratory methods. In particular, aspects of the present disclosure relate to chromatographic analysis methods and systems, as well as chromatographic data analysis methods and systems, such as for monitoring and controlling chromatographic processes and systems.

Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the art to which this invention pertains. The materials, methods and examples are illustrative only and are not intended to be limiting. One of ordinary skill in the art will recognize that routine variations of the disclosed materials, methods and examples are possible without undue experimentation. All publications, patent applications, patents, sequences, databases and other references mentioned herein are incorporated herein by reference in their entirety. In case of conflict, the present specification of the invention, including definitions, will control.

As used herein, the terms comprise, comprising, or any derivative thereof are intended to cover a non-exclusive inclusion, such that a process, method, article, or device that contains a list of elements does not include only those elements but may include other elements not expressly listed or not inherent in such process, method, article, or device. The term exemplary is used in the sense of an example rather than an ideal. For such terms, and for the terms for example and such as and their grammatical equivalents, the expression and without limitation are to be understood as following unless expressly stated otherwise. As used herein, the term approximately and the notation ~ are intended to account for variations due to experimental error. All measurements presented herein are understood to be modified by the term approximately whether or not the term is explicitly used, unless expressly stated otherwise. As used herein, the singular forms a, an, and the include plural references unless the context clearly dictates otherwise. Moreover, in the claims, values, limits and/or other ranges mean a value, limit and/or range of ±10%.

As used herein, the term antibody includes antigen-binding molecules as well as antigen-binding fragments of complete antibody molecules. The terms antigen-binding portion of an antibody, antigen-binding fragment of an antibody, and the like as used herein include any naturally occurring, enzymatically produced, synthetic, or genetically engineered polypeptide or glycoprotein that specifically binds to an antigen to form a complex. Antigen-binding fragments of an antibody can be obtained, for example, from complete antibody molecules, using any suitable standard techniques, such as proteolytic cleavage or genetic engineering techniques to produce recombinant structures, involving manipulation and expression of DNA encoding variable and optionally constant domains. Such DNA is known and/or readily available, for example, from commercial sources, DNA libraries (including, for example, antibody phage libraries), or can be synthesized. DNA can be sequenced and manipulated chemically or by molecular biology techniques, for example, to arrange one or more variable and/or constant domains in a suitable configuration or to introduce codons, incorporate cysteine residues, modify, add or delete amino acids, and so on.

Non-limiting examples of antigen-binding fragments include: (i) Fab fragments; (ii) F(ab') ₂ fragments; (iii) Fd fragments; (iv) Fv fragments; (v) single-chain Fv (scFv) molecules; (vi) dAb fragments; and (vii) minimal recognition units consisting of amino acid residues that mimic hypervariable regions of an antibody (e.g., a single complementarity determining region (CDR), such as a CDR3 peptide), or FR3-CDR3-FR4 peptides with limited conformational freedom. Other genetically engineered molecules, such as domain-specific antibodies, single domain antibodies, domain-deleted antibodies, chimeric antibodies, CDR-grafted antibodies, diabodies, triabodies, tetrabodies, minibodies, nanobodies (e.g., monovalent nanobodies, bivalent nanobodies, etc.), modular small molecular weight protein (SMIP) immunopharmaceuticals, and shark IgNAR variable domains, are also included within the concept of antigen-binding fragment as used herein.

As used herein, the term biological product may refer to a large molecule (e.g., greater than 30 kDa in size) produced in a living system, such as a cell. Biological products may include proteins (e.g., antibodies), nucleic acids, large sugars, etc. Unlike small molecules, which may have well-defined chemical structures, biological products may have very complex structures that cannot be easily quantified by laboratory methods. Thus, it may be desirable to achieve purity, integrity, and quality in the production of biological products to ensure the quality of biological products, particularly when they are intended for medical use.

- 4 047516

As used herein, the term chromatography may refer to any preparative or analytical chromatographic method. Although much of the present invention is presented in the context of preparative packed bed chromatography for the purification of a biological product, it is contemplated that the systems and methods disclosed herein may be applied to a wide variety of chromatographic processes.

As used herein, the term drug product may refer to a volume of formulated drug substance dispensed in a primary packaging component for packaging, transport, delivery, and/or administration to a patient. Drug products may contain active ingredients, including, for example, biological agents.

As used herein, the term crude material(s) may refer to a mixture comprising one or more biological products that require separation or purification by a chromatographic process.

As used herein, the term raw chromatographic data may refer to chromatographic data in the original data state when it was originally collected. For example, raw chromatographic data may be in a .RES file format, another raw file format, or in a database containing values obtained directly from the measurement equipment.

As used here, the term extracted chromatographic data can refer to chromatographic data that has been extracted from raw data without any transformation. This could be an Excel or CSV file, or a database residing on a chromatographic system or computer system.

As used here, the term noise-reduced data can refer to chromatographic data, such as transition data, that have been normalized, smoothed, derived, and/or selected.

As discussed above, there is a need to monitor and control chromatographic column and process quality, such as during multiple chromatographic runs, for multiple batches, and over time both during and between runs. The systems and methods disclosed herein can provide for analysis of chromatographic transition data (also known as transition analysis) and use such analyses to monitor the performance of a chromatographic system, detect changes in the performance of a chromatographic system, and take actions with respect to a chromatographic system based on such analyses and processes. In addition, the systems and methods disclosed herein, in some aspects, can represent part of one or more in-process manufacturing or purification controls and/or can provide in-process control using data collected during routine chromatographic processes, thereby minimizing the increase in cost and labor required to implement separate control processes.

Reference will now be made in detail to the exemplary embodiments of the present invention described below and illustrated in the accompanying drawings. Wherever possible, the same reference numbers will be used throughout the drawings to refer to the same or similar elements.

Fig. 1 is a schematic representation of an exemplary chromatographic system 100 with which various embodiments of the present disclosure may be performed. The system 100 includes a liquid mobile phase feed system 102, a material injection system 104, a column 106, a process controller 108, a computing device 110, and a detector 112.

The system 100 may represent all or part of a chromatographic system, including the chromatographic column 106. In some cases, the system 100 may represent a chromatographic unit. The system 100 may include any hardware and/or software necessary for analysis on a chromatographic column. The system 100 may be configured to perform any of various types of chromatography, such as high performance liquid chromatography (HPLC), ion exchange chromatography, size exclusion chromatography, hydrophobic interaction chromatography (HIC), reversed phase chromatography, mixed mode chromatography, or affinity chromatography. The system 100 may be used, for example, to separate biological products in a crude mixture, to isolate and/or purify one type of biological product, and/or to remove contaminants from the mixture. In some cases, the system 100 may be part of a drug manufacturing system, such as a drug manufacturing system containing a biological product, such as an antibody.

The liquid mobile phase feed system 102 may be any suitable system for feeding a mobile phase to the inlet of the column 106. The liquid mobile phase feed system 102 may include one or more reservoirs for storing the liquid mobile phase(s) used to conduct the raw materials injected by the material injection system 104 through the column 106. The liquid mobile phase feed system 102 may include one or more pumps configured to apply pressure to the liquid mobile phase(s). In some embodiments

- 5 047516 embodiments, the pumps of the liquid mobile phase supply system 102 can be configured to mix two or more solvents (for example, from two or more reservoirs) in a desired ratio before supplying the combined solution to the inlet of the column 106. In some embodiments, the liquid mobile phase supply system 102 can be configured to supply a first mobile phase to the inlet of the column 106 and then supply a second mobile phase to the inlet of the column 106 after the desired volume of the first mobile phase has been supplied. In some embodiments, the liquid mobile phase supply system can be controlled by the process controller 108 or by human intervention.

The material injection system 104 may be any suitable system for feeding the raw material that requires separation and/or purification to the column 106. In some embodiments, for example, the material injection system 104 may include one or more reservoirs for storing the raw material. Such raw material may include one or more biological agents, contaminants, solvents, or other materials.

Column 106 may be any column suitable for separating and purifying the injected raw material from the material injection system 104. One skilled in the art will appreciate that column 106 may have any of a variety of different sizes (e.g., diameters in the range of about 30 cm to about 1500 cm) and may be packed with any of a variety of stationary phases. The size, shape, and packing of column 106 may be selected based on the raw material that requires separation on column 106.

The process controller 108 and/or the computing device 110 may be suitable for controlling the elements of the system 100 during the chromatographic analysis. The process controller 108 may be coupled to one or more elements of the system 100, including the liquid mobile phase supply system 102, the material injection system 104, the column 106, the computing device 110, and the detector 112. In some embodiments, the process controller 108 may be a computer programmed to control the elements of the system 100 in accordance with a desired procedure. For example, in some embodiments, the process controller may be programmed to turn on and off the pumps of the liquid mobile phase supply system 102 and to turn on and off the detector 112. In some embodiments, the process controller 108 may have a display and/or other user interface elements (e.g., buttons, a mouse, a keyboard, a touch screen, etc.) through which commands can be entered, for example, by a human operator. In still other embodiments, process controller 108 may be programmed using, for example, computer device 110.

The computing device 110 may represent any computer, such as a desktop computer, a server computer, a laptop, a tablet, or a personal portable device (e.g., a smartphone). In some embodiments, the computing device 110 may have a display and/or other user interface elements (e.g., buttons, a mouse, a keyboard, a touch screen, etc.) through which commands may be entered, such as by an operator. The computing device 110 may also collect data from the process controller 108 and/or other elements of the system 100, such as the detector 112. The computing device 110 may include one or more programs configured to display or output such data, such as on a screen, a hard drive, or via an Internet connection to a remote location. The computing device 110 itself may be connected to other elements of the system 100 via a wired connection or may be wirelessly connected to other elements of the system 100 (e.g., the process controller 108). In some embodiments, the computing device 110 may be located remotely relative to the system 100. In some embodiments, the computing device 110 may be configured to display one or more user interfaces or reports. In some embodiments, the process controller 108 and the computing device 110 may represent a single device.

The detector 112 may be any type of detector suitable for detecting one or more parameters at the outlet of the column 106. Although one detector 112 is shown in Fig. 1, the system 100 may include more than one such detector configured to detect a plurality of parameters at the outlet of the column 106. Such parameters may include, for example, electrical conductivity at the outlet of the column, pH, optical density, and other parameters. In some embodiments, the detector 112 may be, for example, a conductivity detector, an ultraviolet (UV) detector, a fluorescence detector, a refractive index detector, a pH detector, a manometer, or any other type of detector.

A chromatographic run, such as an analysis performed using the system 100, may typically include a series of steps. Such steps may include, for example, an in-situ cleanup step, an equilibration step, a loading step, a washing step, an elution step, an extraction step, and a regeneration step. The chromatographic run may be monitored and/or recorded using data collected from a detector at the outlet of the chromatographic column (e.g., detector 112 at the outlet of column 106). For example, UV detection and a UV chromatogram may be used

- 6 047516 be used to monitor the chromatography process, such as during the wash, elution, collection and recovery steps. Figure 2 shows an exemplary chromatogram obtained with UV detection, having a typical profile for the collection of a single protein. As the bulk liquid passes through the column (shown on the X-axis), the UV detector detects a clearly steady increase in absorbance with a single peak near the beginning of the elution step. Collection can be started after a small elution peak, during which the absorbance peaks when most of the analyte passes the UV detector.

A chromatographic transition is a response at the outlet of a column (e.g., column 106) to a change in a step at the inlet of the column (e.g., a transition from a wash step to an elution step, or a transition from an elution step to an extraction step) when one mobile phase is replaced by another. Depending on which parameters are detected at the outlet of the column (e.g., by one or more detectors 112), the transition may be detected as an increase (an up-transition) or a decrease (a down-transition) in one or more parameters, followed by a plateau of that parameter after the transition. For example, Fig. 3 shows an exemplary normalized plot of a chromatographic up-transition divided into three phases. Before the transition, the detector detects a value of the parameter at the baseline. During the transition, the parameter increases or increases, and then plateaus after the transition. In some cases, the plateau after the up-transition is due to detector saturation. The data obtained during the transition are quantitative and sensitive to small changes in column efficiency.

Examples of measurable parameters that may change during the transition include electrical conductivity, pH, salt concentration, light absorption, fluorescence after excitation with light of a suitable wavelength, refractive index, electrochemical response, and data obtained by mass spectrometric analysis. One skilled in the art will understand, however, that any other measurable parameters that may change during the transition may be useful in the transition assays of the present disclosure.

In order to perform transition analysis to determine the quality and/or integrity of a chromatographic column and/or process, the chromatography data can be divided into a plurality of blocks, where each block corresponds to a step in the chromatography process (e.g., a clean-in-place block, an equilibration block, a loading block, a washing block, an elution block, an extraction block, a regeneration block, a storage block, etc.). Each block includes a plurality of signals provided by one or more detectors during the block. Any number or combination of blocks and signals can be used to perform transition analysis, such as 1 to 8 blocks (e.g., 1 block, 2 blocks, 3 blocks, 4 blocks, or 5 blocks) and 1 to 8 signals (e.g., 1 signal, 2 signals, 3 signals, 4 signals, 5 signals, 6 signals, or 7 signals). More blocks and/or signals can also be used.

Figure 4 shows an example plot of the detected conductivity as a function of volume during the upward transitions in the equilibration blocks for three chromatographic runs. Each run involved the same chromatography run on the same feedstock on the same column, including the isolation of the same protein, but different batches of feedstock were used. The first peak (in all three runs) represents the baseline of the system. As can be seen, after the peak appears, there is variation in the phase of the transition in the three runs. The shortest dashed line shows the closest to the ideal phase of the transition, since the transition is the most vertical (i.e., occurs in the shortest volume). The longer dashed line shows some indicators of column failure, namely an early start of the transition phase and a tapered end. In general, this transition occurs in a larger volume. The solid line shows more pronounced indicators of column failure, since the transition phase starts very early and takes too long to reach saturation. Although these differences are obvious and visual, they cannot be easily quantified or given context without being compared to each other. The present disclosure provides systems and methods for performing analysis using this data and for reliably performing process control using such analysis.

Figs. 5 and 6 illustrate exemplary processes for analyzing chromatographic data and using such analyses to perform process control according to certain aspects of the present disclosure. Fig. 5 illustrates an exemplary process at a more general level of detail, while Fig. 6 illustrates more details of an exemplary process. Although described separately below, the details and features of the process of Fig. 6 apply to the process of Fig. 5 and vice versa.

Fig. 5 shows an exemplary general process 500 for analyzing chromatographic data and performing process controls according to some aspects of the present disclosure. According to step 510, raw chromatographic data may be processed. According to step 520, data may be obtained from the raw chromatographic data. According to step 530, the obtained data may be processed. According to step 540, the processed data may be analyzed (e.g., transition analysis). According to step 550, one or more statistical process controls may be performed. According to step 560, based on

- 7 047516 all data can be reported.

According to step 510, the raw chromatographic data can be processed. The raw chromatographic data can be obtained by performing one or more chromatographic runs and obtaining signals from one or more detectors (for example, detector 112 of column 106). The signals can include, for example, a UV detection signal, a conductivity signal, a pressure signal, a pH signal and/or other signals. The data can be obtained, for example, in the process controller 108 and/or the computing device 110 and can be stored, for example, in a database or a file of .RES format. The data can include, for example, a series of signal values and the corresponding volumes at which the signal values were measured. The data can also include indicators of the beginning and end of each block/step in the chromatographic run.

The data processing may include data extraction and data organization in a data file in a computing device, such as computing device 110. Exemplary data files include, for example, a spreadsheet, a text file, a database, combinations thereof, and the like. Various metadata may be assigned to the data files containing the extracted chromatographic data to ensure consistent storage and processing. The metadata may include, for example, titles, names, dates, column analysis times, column analysis volumes, column mobile phases, identification of crude mixtures, identification of manufacturing processes for which the column is used, and/or any other data that can ensure consistent automatic or manual processing of the data files.

According to step 520, data can be obtained for analysis from data files. In some embodiments, automatic software (such as Cron, Jobber, macro or other software for automation or sequencing of programs) can monitor one or more possible storage locations of data files for one or more data files that correspond to one or more profiles. The data files can be assigned a profile based on, for example, metadata associated with the data files. The profile can be, for example, a pre-compiled series of selection criteria so that one or more combinations of blocks and signals suitable for performing a transition analysis can be selected. The profile can be assigned based on, for example, the type of column on which the analysis is performed, the characteristics of the mobile phase, the volume of the mobile phase in the analysis, the analysis time on the column or any other characteristics of the data files.

Obtaining data for analysis may include selecting one or more combinations of blocks and signals based on an assigned profile, where the block corresponds to a step in the chromatography process, and the signal corresponds to the type of data to be collected (e.g., UV absorbance data, conductivity, pH, etc.). In some embodiments, a primary combination of blocks and signals may be selected. In additional embodiments, a primary combination of blocks and signals and one or more secondary combinations of blocks and signals may be selected. The transition analysis may be performed first with the primary combination of blocks and signals, and optionally then with one or more secondary combinations of blocks and signals. Profiles, selection criteria, and combinations of blocks and signals are described in more detail with reference to process 600.

According to step 530, the received data can be processed again to obtain data with reduced noise. The processing of the received data can include applying one or more smoothing and/or noise reduction methods to a set of data in the received data, such as data associated with a primary combination of blocks and signals, and, optionally, data associated with a secondary combination of blocks and signals. In some embodiments, the data processing can include standardizing the size of the data set to ensure a consistent effect of the smoothing windows. In some embodiments, the data processing can include normalizing the data to eliminate variation based on the magnitude of the transitions. Such variation can be associated with unique preparations of mobile phase buffers that have their own variability in the final value for the base phase or the saturation phase.

Noise reduction methods may include removing inherent errors introduced by measurement instruments (e.g., detector 112 in system 100) and random errors introduced by periodic processes when data is collected (e.g., in earlier stages of method 500). Noise reduction may include deduplication of records in a data set, detection and removal of outliers, and/or any other method for increasing the signal-to-noise ratio in a data set. Noise reduction may also include data smoothing and signal rejection, which is described in more detail below for process 600.

The processed data may include, for example, the output of the stages and/or measurements of other mobile phase parameters, which may be in the form of one or more smoothed curves corresponding to one or more stages of the chromatographic transitions. The one or more curves may represent an array of normalized data of the resolving signals.

According to step 540, the noise-reduced data may be analyzed. Such analysis may comprise transition analysis. Transition analysis may include performing one or more

- 8 047516 mathematical operations on the processed data. For example, one or more curves can be generated from the processed data, such as by taking the first derivative of the curve to form another curve characterized by a peak. This curve can be analyzed to generate performance parameters, such as the number of inflection points, maximum rate of change, column breakthrough volume, total error, NG-HETP, skewness coefficient, and Gaussian HETP distribution. These performance parameters, alone or in combination with previous data, can help determine the integrity of the column.

For example, an increase in the number of inflection points may indicate that a small amount of early breakthrough is occurring through the column, which may be associated with an integrity problem. A decrease in the maximum rate of change when using multiple columns may indicate that crossover is occurring at a larger volume, which may indicate an integrity problem. A decrease in the volume of breakthrough through the column may also characterize an integrity problem. An increase in NGHETP or the Gaussian HETP distribution may indicate a decrease in column integrity. Other measures of crossover can be derived from the modification of the data set variance, skewness, kurtosis coefficients, peak skewness, breakthrough or elution volume through the column, and total error. Crossover analysis is described in more detail below. Systems and methods for performing crossover analysis are also described, for example, in Larson et al., Use of Process Data To Assess Chromatographic Performance in Production-Scale Protein Purification Columns, 2003, 19, 485-492, which is incorporated by reference in its entirety by pltcm.

The results of the transition analysis may be stored, for example, in a memory element of the computer device 110 or in another computer device together with other data. For example, all raw data, sets of original data, sets of smoothed data and transition analysis data may be stored.

According to step 550, one or more statistical process controls can be performed using the results of the transition analysis. In some embodiments, statistical process control can include performing methods in one of several categories, including 1) abnormal control chart analysis (e.g., analysis of a control chart with individual values, a moving range control chart, and/or a range control chart), 2) multivariate analysis (MVA), or 3) a combination of abnormal control chart analysis and MVA. These processes can include, for example, analyzing the results of the transition analysis as part of a larger data set, including the results of the transition analysis from previous chromatographic analyses, such as analyses in the same production run, analyses of the same product lot, or analyses of the same raw material mixture. These processes are described with additional specifics below for process 600.

The result of performing one or more statistical process controls may be related to performance evaluation data. Performance evaluation data may be related to any process data, including the results of transition analysis, that are relevant in assessing the reproducibility and effectiveness of the process.

According to step 560, the data may be converted into reports. In some embodiments, one or more reports may be obtained. For example, the disclosed methods and systems may generate a tabular report of any results analyzed using a given profile. The reports may be obtained based on a desired number of previous chromatographic analyses, for a certain period of time, for certain analyses, and/or for certain batches. An example of a report is shown in Fig. 24, and is described in more detail below.

Fig. 6 shows, in more detail than Fig. 5, an example process 600 for analyzing chromatographic data and performing process control in accordance with some aspects of the present disclosure. According to step 610, raw chromatographic data can be obtained. According to step 620, the raw chromatographic data can be processed according to a profile. According to step 630, a noise reduction method can be applied. According to step 640, a transition analysis can be performed based on the processed chromatographic data to generate transition data representing the integrity of the column. According to step 650, at least one of an individual value (I) control chart, a moving range (MR) control chart, or a range (R) control chart can be generated based on the transition data. According to step 660, one or more multivariate statistical analysis methods can be applied to at least one I-map, MR-map, or R-map to generate performance data. According to step 660, an action can be performed based on the performance data.

According to step 610, raw chromatographic data may be obtained. As with process 500, the raw chromatographic data may be obtained, for example, from a chromatographic system, such as system 100. The raw chromatographic data may include a plurality of signals associated with a plurality of units. Obtaining the raw chromatographic data may include directly extracting the raw chromatographic data from one or more detectors (e.g., detector 112 of system 100) or from a

- 9 047516 computer device (e.g., computer device 110), and/or may include monitoring a network location of a raw chromatographic data file. The raw chromatographic data may, in some embodiments, be processed as described above for step 510 in process 500.

An exemplary extracted chromatographic data file 1000 is shown in Fig. 7. The data file 1000 may include, for example, a data file name, which may assist in identifying the data file using an automated system. As shown, the extracted chromatographic data in the data file 1000 may be in the form of a spreadsheet (e.g., Microsoft Excel). The data file 1000 may include volume measurements on a first column 1002, which may correspond to periodic measurements of the total volume that has passed through the chromatographic system. A second column 1004 may include signal measurements (e.g., UV absorbance, conductivity, pH, etc.) corresponding to each of the volume measurements on the column 1002. In this case, the second column 1004 contains conductivity data expressed in mS/cm. Other columns may provide additional data. Here, for example, the third column 1006 includes volume measurements corresponding to the logs in the fourth column 1008. This may allow for the identification of various chromatographic analysis parameters, such as the start and end points of a block/step (CG002_START, CG002_END, CG003_START), flow rate, and points at which elements of the chromatographic system were initiated (e.g., pump 1 may correspond to the time when a pump, for example, associated with the liquid mobile phase supply system 102, is activated). One skilled in the art will appreciate that many variations of the data file 1000 are possible. For example, although volume measurements are shown in the data file 1000 as markers of the progress of the chromatographic analysis, other measurements, such as time, may be used. Additional columns for other signal data may be included, and additional log data may be included (e.g., mobile phase identification, analyte identification, etc.).

Referring to Fig. 6, according to step 620, the chromatographic data can be processed according to a profile. As briefly described for step 520, a profile can be selected for a chromatographic data file according to the characteristics of the chromatographic data in the file. For example, profiles can be previously created for a specific type of chromatographic analysis, a specific chromatographic column and/or a specific analyte. In this way, such profiles can be compared with a chromatographic data file for the corresponding analysis, column and/or analyte.

In some aspects, the profile may be created by a user. The profile may be associated with a particular drug or drug product. In one aspect, the drug is a small molecule. In other aspects, the drug is a peptide or polypeptide.

In some aspects, the drug is a vascular endothelial growth factor (VEGF) derivative. In other aspects, the drug is aflibercept, which is described in one or more of U.S. Patent Nos. 7,070,959, 7,303,746, 7,303,747, 7,306,799, 7,374,757, 7,374,758, 7,531,173, 7,608,261, 7,972,598, 8,029,791, 8,092,803, 8,343,737, and 8,647,842, each of which is incorporated herein by reference in its entirety.

In still other aspects, the medicament is an antigen-binding molecule. In some aspects, the antigen-binding molecule is an antibody or an antigen-binding fragment. In some aspects, the medicament is alirocumab, which is described in U.S. Patent Application Publication Nos. 2014/0356371 and 2014/035670, each of which is incorporated herein by reference in its entirety. In another aspect, the medicament is sarilumab, which is described in U.S. Patent Application Publication Nos. 2016/0152717, 2014/0302053, and 2013/0149310, each of which is incorporated herein by reference in its entirety. In yet another aspect, the medicament is dupilumab, which is described in U.S. Patent Application Publication No. 2014/0356372, which is incorporated herein by reference in its entirety. In another aspect, the drug is selected from the group consisting of evolocumab, bevacizumab, ranibizumab, tocilizumab, certolizumab, etanercept, adalimumab, abatacept, infliximab, rituximab, anakinra, trastuzumab, pegfilgrastim, interferon beta-1a, interferon beta-1, interferon beta-1, insulin glargine (rDNA) injectable, epoetin alfa, darbepoetin, filigrastim, and golimumab.

In some embodiments, the profile may be configured to direct Sentinel software (e.g., a macro, Jobber, Cron, or other software depending on the programming task) to periodically scan a designated network location for chromatographic data files. The profile may direct data retrieval from a file when the file name matches a file name identifier in the profile.

Once a profile is or has been selected on or associated with a data file, the data file can be scanned. For example, with respect to the exemplary data file 1000 shown in Fig. 7, the fourth column 1008 containing the log can be scanned for block start times, block end times, flow rates, and the like. For example, with respect to the data file

- 10 047516

1000, then the volume measurements corresponding to CG002_START and CG002_END include the volume measurements that correspond to the chromatographic operation and the signal transition of interest. The first column 1002 and the second column 1004 can then be used to extract the complete set of signal data and volume measurements for the operation.

The values in the profile may also define one or more selection criteria for selecting one or more combinations of blocks and/or signals in the chromatographic data file for which a transition analysis is to be performed. In this way, profiles may provide tools for obtaining preferred subsets of data from the chromatographic data file. The selection criteria in the profile may be defined in advance, for example, from empirical experience, structured optimization and/or process documentation. Such selection criteria may allow for the identification of combinations of blocks and signals that provide more precise, accurate or otherwise more useful analysis. Such selection criteria may include, for example, how accessible materials are across transitions. This includes blocks that transition to or from product solutions. This allows for additional column evaluations to be performed between production runs, if necessary. Such selection criteria may also or alternatively include how blocks occur at regular intervals in the run. This includes blocks that are not fulfilled after the final collection run of the production batch has been completed. Such selection criteria may also or alternatively include whether signals reach detector saturation before or after the transition. Such selection criteria may also or alternatively include how much the signals approach a stationary phase at a discernible and identifiable level and do not drift continuously. Such selection criteria may also or alternatively include how much the signals in a given block have a large difference between minimum and maximum values. Such selection criteria may also or alternatively include how many inflection points the signals have during the transition. Fewer inflection points may indicate more reliable data collection.

In some cases, the results of previous chromatographic analyses can help in determining suitable selection criteria for selecting block and signal combinations in future chromatographic analyses. Figure 8, for example, shows a graph of NG-HETP calculations for two different block and signal combinations (elution step-UV absorbance signal combination and re-equilibration step-conductivity signal combination) for six different chromatographic lots (lots A-F). The solid bars representing three standard deviations for each set are provided as references. As can be seen from this graph, the NG-HETP calculations for the elution step-UV absorbance signal combination show much greater variability than the re-equilibration step-conductivity signal combination. It is clear that both the magnitude of the trends and the standard deviations differ. When monitoring shifts in performance, it may be desirable to have less variability across lots that are considered typical. This allows for increased sensitivity in monitoring shifts in efficiency. Thus, selection criteria for chromatographic analyses of batches similar to batches A-F may include a preference for a combination of a re-equilibration step and a conductivity signal over a combination of an elution step and a UV absorbance signal. One skilled in the art will recognize that analyzing the results of previous chromatographic analyses in a similar manner may reveal other potential selection criteria for combinations of blocks and signals.

In some embodiments, the profile may include instructions for applying one or more selection criteria to a data file containing corresponding chromatographic data. Thus, processing the chromatographic data according to the profile may include identifying and retrieving a preferred (e.g., primary) combination of blocks and signals for transition analysis and/or one or more additional (e.g., secondary) combinations of blocks and signals for transition analysis. In some embodiments, the primary combination of blocks and signals will match the majority of the selection criteria in the profile from all possible combinations of blocks and signals in the chromatographic data file. In some embodiments, the secondary combination of blocks and signals will match the second selection criterion in the profile from all possible combinations of blocks and signals in the chromatographic data file. Although the primary combination of blocks and signals may include data that is likely to provide useful transition analysis for assessing column and process integrity, the secondary combination of blocks and signals can provide a secondary measurement and cross-check of column integrity.

In some embodiments, the profile according to step 620 may be a data file in or by itself, which may contain instructions for extracting certain data from a chromatographic data file or changing a chromatographic data file with corresponding metadata. In some embodiments, such instructions in the profile may be executed by a computer program.

- 11 047516

Referring to Fig. 6, after the chromatographic data has been processed according to the profile, a noise reduction method can be applied to the processed data according to step 630. As in the case of step 530 of the process 500, this step can include applying one or more smoothing and/or noise reduction methods to the processed data (e.g., data associated with the selected combinations of blocks and signals). In some embodiments, this step includes standardizing the size of the data set to ensure a consistent effect of the smoothing windows. In some embodiments, this step can include normalizing the data to eliminate variation based on the magnitude of the transitions. Such variation can be associated with unique compositions of the mobile phase buffers, which have their own variability in the final value for the base phase or the saturation phase.

Noise reduction methods may include removing inherent errors introduced by measurement instruments (e.g., detector 112 in system 100) and random errors introduced by periodic processes when data is collected (e.g., in earlier stages of method 500). Noise reduction may include deduplication of records in a data set, detection and removal of outliers, and/or any other method for increasing the signal-to-noise ratio in a data set.

Noise reduction may also or alternatively include applying a data smoothing and signal error rejection algorithm. In this regard, an exemplary algorithm 900 is shown in flow chart form in Fig. 9. According to steps 902 and 904 of the algorithm 900, the algorithm may be run and the corresponding signal data (e.g., the data that was processed according to step 620) is extracted. According to step 906, the extracted data may be normalized to remove the amount of rejection.

Then, a multi-level smoothing algorithm 950 can be applied. This can include applying one or more initial smoothing filters (steps 908, 910) in accordance with the desired settings of the smoothing filters (909, 911). According to step 912, derivation can optionally be performed. Then, one or more additional smoothing filters (steps 914, 916) can be applied in accordance with the desired settings of the additional smoothing filters (913, 915). The number of smoothing filters applied (steps 908, 910, 914, 916), as well as the number and values of the settings 909, 911, 913, 915 can vary depending, for example, on the state of the data, the expected results, the type of signals and other factors. Whether data derivation will be performed can also depend on these factors.

The process may then continue with the dynamic signal error rejection algorithm 980. This algorithm may be configured to remove data from the extracted data that is not related to the chromatographic transition. For example, errors that must be removed to ensure a meaningful transition analysis include alarms, machine stops, setup sensor failures, or data gaps. This may be achieved by determining expected chromatographic transition indices, such as the derivative duration; maximum intensity; duration from initiation; or expected background noise. For example, the starting point 918 rejection may be made based on the expected transition position 919, the initial dead zone 920 rejection may be made based on the expected background noise level 921, the height and width derivative rejection may be made based on the expected signal error indices, and the final dead zone rejection may be made based on the expected background noise levels 925. The expected transition indices may be generated, for example, based on previously accumulated transition data. After completion of algorithm 900, in accordance with stage 990, the data may be ready for use in transition analysis.

Although the algorithm 900 is one example model of the algorithm for smoothing and rejecting the signal error, a person skilled in the art understands that variations of this algorithm are possible. For example, only the smoothing algorithm 950 can be performed, or only the signal error rejection algorithm 980 can be performed. Additionally or alternatively, more or fewer smoothing filters can be applied, and/or more or fewer points can be rejected.

After applying a noise reduction and/or smoothing method to the data, the data may include, for example, stage yields and measurements of other mobile phase parameters in the form of a breakthrough or washout curve corresponding to the stage transition.

Referring to Fig. 6, at step 640, a transition analysis may be performed on the processed chromatographic data to obtain transition data representative of the integrity of the column. The transition analysis may include performing one or more mathematical operations on the processed data to derive dispersion parameters from the step transition. For example, one or more curves may be constructed from the processed data, such as by taking the first derivative of the curve to form another curve characterized by a peak. This curve may be used to generate performance parameters such as, for example, the number of inflection points, the maximum rate of change, the breakthrough volume through the column, the total error, NG-HETP, the skewness coefficient, and the Gaussian HETP distribution. As described for step 540, these parameters may be used

- 12 047516 be used as indicators of column integrity or lack thereof (e.g. when compared with transition analysis parameters from previous representative chromatographic data).

For example, an increase in the number of inflection points may indicate that a small amount of early breakthrough is occurring through the column, which may be due to a loss of integrity. The number of inflection points can be determined by the number of peaks when plotting the derivative curve against the cumulative volume data.

As another example, a decrease in the maximum rate of change when multiple columns are used may indicate that the transition is occurring in a larger volume, which may be an indication of a breach in integrity. The maximum rate of change is equivalent to the maximum value of the derivative curve.

As another example, a decrease in column breakthrough volume can also indicate a loss of integrity. Column breakthrough volume can be determined by finding the first volume value where the signal is less than 95% of its highest value (for a high-to-low transition) or greater than 5% of its lowest value (for a low-to-high transition).

As another example, an increase in NG-HETP or the Gaussian HETP distribution may indicate a decrease in column integrity. Other transition indices may be derived from the modification of the data set variance, skewness, kurtosis, peak skewness, column breakthrough or elution volume, and total error. Systems and methods for performing transition analysis are also described, for example, in Larson et al., Use of Process Data To Assess Chromatographic Performance in Production-Scale Protein Purification Columns, Biotechnol. Prog., 2003, 19, 485-492 2003, 19, 485-492, which is incorporated herein by reference in its entirety.

The results of the transition analysis may be stored, for example, in a memory element of the computer device 110 or in another computer device together with other data. For example, all raw data, sets of original data, sets of smoothed data, and transition analysis data may be stored.

Referring to Fig. 6, according to step 650, at least one of an individual value (I) control chart, a moving range (MR) control chart, or a range (R) control chart can be obtained based on the transition data. For simplicity, the present disclosure will refer to all of these collectively as an I-MR-R chart; however, an I-MR-R chart should be understood to refer only to an I chart, only to an MR chart, only to an R chart, or any combination and number of such charts. The I-MR-R control chart is an individual visualization of the transition analysis results and can assist in interpreting trends in the transition analysis data across multiple column or batch analyses in the form of NG-HETP, skewness, kurtosis, or other parameters. An advantage of I-MR-R charts is that the data can be quickly viewed and easily interpreted from a visual point of view. This makes small trends or immediate changes in the data recognizable at an early stage.

An I control chart, for example, may display a plot of values for each lot analyzed (e.g., a skewness ratio). An MR control chart may display a plot of values for the difference between the value for each lot analyzed and a previously analyzed lot. An R control chart may display a plot of values for the difference between values within a lot (e.g., a skewness ratio for two transition analyses performed on a single lot, for a primary combination of blocks and signals and a secondary combination of blocks and signals). Each control chart may include a median line, upper control limits (UCLs), and lower control limits (LCLs), which may be calculated using available data that have been determined to be representative of a typical process, and are located at equal distances from the median line on each control chart.

Some parameters presented on I-MR-R control charts, such as NG-HETP and skewness coefficient in transition analyses, may show significant dynamics over the lifetime of certain limits. In such cases, using an I-MR-R chart with control limits established using a short-term standard may result in excessive off-trend signals, even after resetting the chart following column repacking. One solution to this problem is to use a Levey-Jennings control chart, which uses long-term standard deviation calculations from representative column lots that take into account the unique variations associated with starting a new column pack. How representative the data are can be determined by the absence of outliers across the various performance evaluation data sets for the lot. These sets can be used to calculate the standard deviation, sometimes with special attention to the ±3 standard deviation (SD) lines. Multiple lots can be analyzed on a single column to determine whether the column's primary or full life was typical. In one aspect, a full life-cycle modeling of the column can be performed on the Levey-Jennings control chart, resulting in a regression model that accounts for the special causes of variability when repacking the column. However, for the Levey-Jennings control chart, a full life-cycle modeling of the column can be performed, resulting in a regression model that accounts for the special causes of variability when repacking the column. However, for the Levey-Jennings control chart, a full life-cycle modeling of the column can be performed.

- 13 047516 Levey-Jennings trace map requires longer term data and therefore its use will be limited by the speed of data aggregation.

Additionally, since transition analysis is known to have variations due to column repacking events, I-MR-R control charts can account for column packing and repacking - for example, the first analysis of a lot after a column repacking will not have an MR value based on the variation of the last analysis of the lot before the column repacking. In some respects, control strategies can be configured to only consider specific violations that eliminate the known variation due to repacking events when monitoring for deviations in trends.

The generation of I-MR-R control charts may be performed, for example, by an assay module in the computer device 110 or another assay module elsewhere. The generation of the I-MR-R chart may also be performed in the computer device 110, for example, by a control chart module. For example, Figs. 14-21 show I-MR-R data for chromatographic lots 21-100 and are discussed below.

Referring again to Fig. 6, according to step 660, one or more multivariate statistical analysis methods can be applied to the I-MR-R data. Alternatively, one or more multivariate statistical analysis methods can be applied to the transition analysis data. This step can be performed in addition to or as an alternative to step 650 and similar to generating maps according to step 650, taking into account the transition analysis of previous chromatographic data. Multivariate statistical analysis takes numerous variables and simplifies them into component vectors. This allows for a holistic view of large data sets. Advantages include that numerous, subtle changes in multiple executions that would not be apparent when looking at individual data sets may become apparent when displaying their component vectors. Variations in these data may be caused by differences in materials, equipment, ambient atmospheric conditions, etc., and may be insignificant to the perception of an operator or human observer. Examples of multivariate statistical analysis methods may include principal component analysis (PCA), partial least squares (PLS), orthogonal partial least squares (OPLS), multivariate regression, canonical correlation analysis, factor analysis, cluster analysis, graphical procedures, and the like. Such multivariate statistical analyses may be performed using, for example, specialized computer software.

The overall goal of using multivariate analysis is to transform large amounts of data into interpretable information. By being able to find correlations and patterns among multivariate variables and extract statistically significant values from large amounts of raw data, multivariate analysis allows for the interpretation of, for example, any statistical significance for the variation between analyses of similar batches of chromatographic data.

For example, PCA is a multivariate statistical technique in which a data set containing many variables (e.g., a transition analysis involving several parameters) is reduced to a few variables, called t-scores. For example, a data set containing many variables can be reduced to a data set where each observation (e.g., each transition analysis) is represented by two t-scores. T-scores contain information about the variation of each variable in the data set and the correlation of each variable with every other variable in the data set. Essentially, t-scores describe the pattern of variation and correlation of each observation (e.g., each transition analysis) in the data set to every other observation in the data set. A graphical representation of the results of PCA is usually a PCA plot. A PCA plot is a plot of one t-score against another for each observation. Typically, a PCA plot is a distribution that shows how the pattern of variation and correlations compares for all the observations in the data set. In this way, the plot can serve to group similar observations together.

As another example, PLS regression analysis is a method for analyzing systems of independent and response variables. PLS is a predictive method that can handle many independent variables, even if the variables are multicollinear. PLS can also relate a set of independent variables to a set of multiple dependent (response) variables. Often in PLS, one set of latent variables may be extracted for a set of manifest independent variables, and another set of latent variables may be extracted for a set of manifest response (or dependent) variables. This extraction process may be based on the decomposition of a cross-product matrix including both the independent and response variables. The scores or x-values of the latent independent variables are used to predict the scores or y-values of the latent response variables. The predicted y-values are then used to predict additional response variables. The x and y scores are chosen such that successive pairs of x- and y-variables are as closely related as possible. The advantages of PLS include the ability to model multiple

- 14 047516 independent and dependent variables, the ability to control for multicollinearity between independent variables, robustness to data noise and (depending on the software used) missing data, and to create independent latent variables directly from cross-products involving the response variable(s), making it possible to obtain more accurate forecasts.

In some embodiments, multivariate statistical analysis may be performed on the I-MR-R control chart to determine additional statistical significance of the variation shown on the I-MR-R chart.

In addition to the analyses described above, trends in transition analysis can be created by calculating non-stationary ranges that allow slow variations to remain within control limits, while abrupt shifts in column efficiency can be flagged as off-trend potentials. The main methods for determining control limits include moving averages, weighted moving averages, and varying degrees of exponential smoothing. One such method for calculating trend limits, known as the Holt-Winters or triple exponential smoothing method, can be used with high efficiency. The Holt-Winters method uses seasonality to predict appropriate limits, which are defined as individual column packing events for direct application to chromatographic monitoring. Regression modeling (such as that used in the Levey-Jennings control chart) is an additional way to establish trend limits. Once sufficient empirical data are available, regression modeling of column integrity can be performed on the total column packing usage. This can provide accurate, consistent column efficiency ranges based on historical column efficiency data included in the model.

Referring again to Fig. 6, according to step 670, an action can be performed based on the performance data. In some embodiments, this can be due to the fact that the transition analysis is identified as an in-process control (IPC). In general, the action of step 670 can include generating a report, generating and/or transmitting an alarm signal to an operator or to a display, such as a display of the computer device 110, or ending the chromatography process. The action of step 660 can also include, for example, storing all data obtained during the systems and methods disclosed herein in a database for further analysis.

The result of performing a multivariate analysis and/or an I-MR-R control chart analysis on transition data may be related to performance evaluation data. Performance evaluation data may be related to any process data, including the results of a transition analysis, that may be relevant in assessing the reproducibility and performance of a process (e.g., a chromatography process).

In one aspect, step 670 may include generating one or more reports. For example, the disclosed methods and systems may generate reports in a tabular format of any results analyzed using a given profile. The reports may be created based on an analysis of a desired number of previous batches, for a specific time period, and/or for specific batches. The data sets may be fully extracted in multiple formats and may be input into external applications if further analysis is required.

Fig. 24 is an example report 2400 in accordance with some aspects of the present disclosure. The example report 2400 includes a report summary table that includes the results of four chromatography runs from a single manufacturing lot. Each of the four runs is listed by lot and run number and includes a list of the dates and times the analysis was performed. The results of the transition analysis are presented in columns that include NG-HETP, HETP Gaussian Distribution, skewness, skewness, kurtosis, non-Gaussian N, and Gaussian N. A data source snapshot is also provided, indicating the name of the chromatography system from which the data was obtained, the log in which it was recorded, and the blocks for which the data was obtained. Below the data for each of the runs, trend data is provided for each of the analysis results. It should be understood that this report is exemplary and many variations are possible. For example, a desired number of chromatography runs may be provided and/or one or more trend data graphs may be included.

In some aspects, the systems and methods disclosed herein can be used to continuously monitor the integrity of a column and a process. As such, using the systems and methods disclosed herein, data can be analyzed with respect to a particular column and/or process. In one aspect, one or more alarms can be generated based on the analysis of the data. In another aspect, the chromatographic process can be stopped based on the analysis of the data. For example, one or more notifications (e.g., an event notification, an event assessment, or a deviation notification form) can be provided or displayed to the operator for corrective action. For example, one or more object overlays can be displayed on a screen, such as on the screen of the computing device 110, and/or a message can be sent to the operator for a time to complete

- 15 047516 analysis, with a recommendation on whether to continue or stop the chromatography process.

In one aspect, the results of the disclosed methods and systems can be directed to convey information about current trends in the evaluation of the quality of the column packaging before the column is used in production. In another aspect, the results of the disclosed methods and systems can be used to evaluate the performance of the column in real time (or offline) and can confirm the integrity of the column before the next cycle of product use (for example, if the acceptable range and control limits in the trend map are established).

In an additional aspect, the results can be used with statistical information to predict process results based on process modeling using multivariate statistical analysis prior to expensive and time-consuming research and testing.

For example, one evaluation criterion for statistical analysis plots might include that when generating a score plot for a data set using PCA, a lot that is below a threshold number of standard deviations from the mean can be identified as having a column integrity problem and can result in the generation of an alarm or an instruction to change the lot.

In particular, one evaluation criterion for I-MR-R control charts may include that any points outside the upper or lower control limits for one or more chart types may be the basis for an alarm signal. Thus, the action performed according to step 670 may be to issue an alarm signal, for example, from the computer device 110, if the lot shows points outside the control limits. Such alarm signals may include, for example, an event notification, an event evaluation, and/or a deviation notification form, for provision to an operator or to a database.

In some aspects, the systems and methods disclosed herein may be implemented as part of an in-process monitoring system that may operate within an organization's quality control system to ensure consistency and compliance with safety requirements. As part of such a program, data from the systems and methods disclosed herein may be used to determine critical process parameters (CPPs) and critical quality attributes (CQAs) that are to be monitored in the in-process monitoring program. Additionally, as part of such a program, signal transitions and column integrity shifts may be detected in real time or near real time (e.g., during or concurrent with column operation), allowing preventive and corrective actions to be taken in response to performance data.

Fig. 23 shows an exemplary user interface 2300 in accordance with some aspects of the present disclosure. The user interface 2300 depicts a transition analysis profile creation/edit screen by which a user can generate or edit a new transition analysis profile. The parameters selected during profile creation can be used to customize the transition analysis based on the unique characteristics of the chromatographic process and to optimize the robustness of the output data for each column and program. The parameters provided in the exemplary user interface 2300 include, for example, a profile name, comments, historical data and/or test position, file template, end value, key indicator, hard reset, window size for moving average, values for the first filter (e.g., SG filter), values for the second filter, the percentage of the first Vmax value below which the signal should be registered as zero, the percentage of the maximum width for peak retention, the height of the chromatographic column, the start date, the end date, and the name of the database.

The methods and systems disclosed herein can be used to monitor the integrity of a column relatively continuously. For example, the methods and systems disclosed herein can monitor the integrity of a column without having to interrupt regular chromatographic processes in order to perform diagnostics on the chromatographic system. In addition, the methods and systems disclosed herein can analyze data for a specific column and a specific process. As discussed, one or more alarms can be generated based on the analysis of data over time. In another aspect, based on the analysis of data, the chromatographic process can be stopped. For example, one or more notifications can be displayed to an operator to take corrective action in the event that the integrity of the column is compromised. For example, one or more object overlays can be displayed on a screen, and a message box can be displayed to the operator at the time of completion of the analysis, a recommendation to continue or stop the chromatographic process, or a recommendation for other actions.

In some aspects, the results of the disclosed methods and systems can be used to communicate current trends in the evaluation of column packaging quality prior to using the column in production. In other aspects, the results of the disclosed methods and systems can be used to evaluate column performance in real time (or offline) and can confirm the integrity of the column prior to the next product use cycle (e.g., if acceptable

- 16 047516 the range and control limits on the trend map are set). In some aspects, the results can be used with statistical information to predict process results based on process modeling using MVA prior to costly and time-consuming research and testing.

Examples

Example 1.

The primary combination of blocks and signals was selected based on the protein A affinity capture chromatography data as follows. The affinity capture data included eight blocks and two signals (UV absorbance and conductivity) in each block, yielding a total of 16 possible combinations of blocks and signals.

The data is assigned a profile that includes a series of block and signal selection criteria that are applied in the following order to select a primary combination of blocks and signals.

Given the selection criteria that the blocks must appear at regular intervals among production periodic cycles, two blocks and their corresponding signals can be removed, leaving 12 potential combinations.

Considering the selection criteria that the signal must reach saturation of the UV absorbance indicator, the UV signal can be removed for three candidate blocks, leaving nine potential combination options.

Given the selection criteria that signals approach stationary phase at a distinguishable and identifiable level, the UV signal can be removed for three candidate blocks, leaving six potential combinations (all with conductivity as the selection signal).

Given the selection criteria that signals must have a large difference between their minimum and maximum values in a given block, conductivity can be removed for four blocks, leaving two potential combinations.

Given the selection criteria that signals displaying the fewest inflection points are preferred, the conductivity for one block can be removed, leaving only one block and signal combination option.

The final selection of the remaining block and conductivity signal is the primary block and signal combination on which the transition analysis can be performed. The last block and signal combination to be removed becomes the secondary block and signal combination.

Example 2.

The I-MR-R control chart data for the skewness coefficient and NG-HETP trends were plotted for 100 chromatographic batches in this chromatography program B as follows. Figures 14-16 show the I, MR, and R control charts, respectively, showing the skewness coefficient. Figures 17-19 show the I, MR, and R control charts, respectively, showing NG-HETP. The UCL and LCL indicate 3 standard deviations as determined by previously accepted data. Breaks in the mean, UCL, and LCL lines indicate column repacking. Continuous shifts in these lines indicate the point where the limits were recalculated.

Figure 14 shows the skewness coefficient for all 100 batches prepared in program B. It can be seen that the first and second column packs exhibit different behavior during their use. As shown, pack 1 shows a shift in the boundaries after the analysis of the first four batches and maintains the skewness coefficient values from 0.055 to 0.855. Pack 2 is outside this trend but eventually reaches a steady state at batch #67. This may be due to the fact that the shift and settling of the new column pack takes longer than for pack 1.

Fig. 15 shows the MR control chart for the asymmetry coefficient for all lots prepared in program B. For package 2, outliers can be observed indicating large shifts between lots based on individual values.

Figure 16 shows the control R-chart for the skewness coefficient for all lots prepared in program B. Some outliers are noted for pack 1. This has increased the margins for pack 2. There are three packs on the map, and the lots are plotted sequentially so that pack 1 represents the leftmost solid line and pack 3 represents the rightmost solid line. Note that the trend points are missing in the second half of pack 1. This may indicate that the column experienced variation across batch cycles.

Figure 17 shows the control I-chart for NG-HETP for all batches prepared in program B. For pack 1, a decrease in NG-HETP is observed, indicating an improvement in the column behavior model. For pack 2, a continuous increase in NG-HETP is observed, which could correlate with a decrease in column efficiency.

Fig. 18 shows a moving average control chart for NG-HETP for all batches prepared in program B. Outliers can be noted for both package 1 and package 2. This revealed several points showing significant shifts from individual values to

- 17 047516 other individual values.

Figure 19 shows the R-chart for NG-HETP for all batches prepared in program B. Package 2 shows consistently elevated swing values, which were investigated and determined to be due to a flow direction reversal in the third run of the batch. This resulted in the third run showing a different value than the other runs.

Example 3.

Control charts with individual values (I) were plotted on the graph of the transition analyses of two groups of chromatographic batches for a given program A.

Figure 20 shows the control chart I for NG-HETP for 46 batches prepared in program A. The data show that the column is performing within the established limits for process consistency.

Fig. 21 shows the control chart I for NG-HETP for 21 additional lots prepared in program A. The data show that two lots (56 and 58) exceeded the upper control limits.

Example 4.

Multivariate analysis was performed using the transition analysis data for 27 chromatographic lots, including the three lots shown in Fig. 4. Loading values were calculated for seven parameters from the 27 lots, including the three lots shown in Fig. 4. The seven parameters included NG-HETP for each of the control I-chart, control MR-chart, and control R-chart for the lots, the skewness coefficient for each of the control I-chart, control MR-chart, and control R-chart for the lots, and the kurtosis coefficient for the I-chart. Figure 10 shows a plot of the loading of each of the seven parameters. The value of each column corresponds to the effect of the parameter on the principal component. The error bars indicate the relative error in the loading value.

Figure 11 shows an example score plot for 27 lots. The score plot was calculated for seven parameters for the 27 lots, including the loading values calculated for the lots shown in Figure 4 (principal component 1) as well as principal component 2. Lots with similar parameter values were grouped. The ellipse around most points in the plot excludes outliers with a 95% probability.

Example 5.

A multivariate analysis was performed on the I-MR-R control chart data for transition analysis for 46 chromatographic lots as follows. I-MR-R chart data were collected for each of the 46 lots. Lots that were considered atypical or unsuitable based on the I-MR-R chart data were removed from the analysis and the data for the remaining lots are summarized in Table 1 below. Lots containing values for multiple transitions were averaged and presented as individual measurements. Range values were calculated as the maximum values minus the minimum transition values within a lot.

Table 1

Batch number Individual NG-HETP values Individual values of the asymmetry coefficient NGHETP scope Span of the asymmetry coefficient 1 0,0824 0.73 0,01 0.09 2 0,0951667 0.84 0,007 0,07 3 0,0994 0.826 0,007 0,07 4 0,206167 -0.263333 0.204 1.62 5 0.96625 0,135 2,037 1.75 6 1,91875 -0.10625 3,622 1.83 7 0,107 0.925 0,015 0,06 8 0,55075 0,25625 3,355 2.16 9 0.738625 -0.06375 3,418 2.32

- 18 047516

10 0,565714 -0.05 1,302 2.72 And 0,107667 0,715 0,012 0.14 12 0,0745714 0.595714 0,009 oh, and 13 0,0651429 0.56 0,006 0,07 14 0,0575714 0,472857 0,002 0,07 15 0,0575714 0,5 0,007 0.17 16 0,054 0.395714 0,008 0.16 17 0,0701429 0.302857 0,115 0.77 18 0,0628571 0,545714 0,019 0.26 19 0,0702857 0,452857 0,018 0.47 20 0,0671429 0,491429 0,033 0.45 21 0,111429 0,172857 0,06 1,19 22 0,167429 -0.101429 0.163 1.41 27 0,1274 0.544 0,009 0.08 28 0,131833 0.575 0,009 0,06 29 0,136667 0,605 0,009 0,05 31 0,133833 0,6 0,011 0,04 32 0,134833 0.595 0,005 0,03 33 0,1368 0.626 0,003 0,04 34 0,135 0,613333 0,003 0,06 35 0,1344 0,632 0,005 0,05 41 0,137667 0,638333 0,014 0,07 42 0,134833 0.62 0,007 0,05 43 0,1352 0,642 0,004 0,03 44 0,1316 0,638 0,008 0.07 45 0,131833 0,641667 0,003 0,07 46 0,135167 0,675 0,009 0,03

Using the data in Table 1, the principal component was calculated by plotting loading plots showing the coefficients for each input parameter. Each data series was converted to a single value. An assessment of the accuracy and relevance of the model to the physical system was shown by the R2 and Q2 values of the PCA model, where R2 represents a statistic of how close the test data set is to the fitted regression line, and Q2 represents a statistic of how close the test data set should be to the regression line. Together, R2 and Q2 indicate how well the model describes the system being analyzed, with 1 representing a perfect fit and 0 representing no correlation at all.

In Fig. 12, the model loading plot is shown. The R ² value of the model was 0.798 and the Q ² value was 0.591, which indicates that the model was acceptable for application and that all input values had an effect on the principal component of the model since they were not located near the center line. In Fig. 12, the y-coordinate value of each point corresponds to the effect of the parameter (e.g., the effect of the NG-HETP mean, NG-HETP range, skewness coefficient range, and skewness coefficient mean) on the principal component. The y-coordinate of each point corresponds to the number of inputs per point.

The values of the principal components were varied and plotted linearly with respect to the corresponding lots. Figure 13 shows the score plot for the data set. The score plot shows the PC1 value (the value contributed in the direction of the greatest variation) for each lot used. As can be seen from Figure 13, one lot (lot 6) was beyond three standard deviations, and several points were close to exceeding two standard deviations, indicating that the system has variation in these lots.

As will be appreciated by those skilled in the art, the methods and systems disclosed herein may take the form of entirely hardware embodiments, entirely software embodiments, or embodiments combining software and hardware aspects. In addition, the systems and methods of the present invention may take the form of computer program products on a computer-readable storage medium having computer-readable instructions (e.g., computer software) embodied in the storage medium. Suitable computer-readable storage media may include hard disks, CD-ROMs, optical storage devices, or magnetic storage devices. More particularly, the present methods and systems may take the form of Internet-based computer software.

Embodiments of the present invention are described with reference to block diagrams and flow charts of methods, systems, devices and computer program products. It is understood that one or more blocks in the block diagrams and flow charts, respectively, may be implemented by computer program instructions. These computer program instructions may be loaded

- 19 047516 wives in a general-purpose computer, special-purpose computer or other programmable data processing device to obtain a machine such that instructions that are executed on the computer or other programmable data processing device create a means for implementing the functions indicated in the flow chart or blocks.

Such computer program instructions may also be stored in a machine-readable memory that can direct a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the machine-readable memory produce an article of manufacture that includes machine-readable instructions for implementing a function specified in a flow chart or blocks. The computer program instructions may also be loaded into a computer or other programmable data processing device to cause a series of operational steps to be performed on the computer or other programmable device to create a process implemented by the computer, so that the instructions that are executed on the computer or other programmable device provide steps for implementing the functions specified in the flow chart or blocks.

Accordingly, the blocks in the flow charts and process flow diagrams support combinations of means for performing the specified functions, combinations of stages for performing the specified functions, and software instructions for performing the specified functions. It should also be understood that each block in the flow charts and process flow diagrams and combinations of blocks in the flow charts and process flow diagrams can be implemented by hardware of computer systems that perform the specified functions or stages, or by combinations of hardware (for example, special hardware for chromatography) and computer instructions.

Fig. 22 shows an operating environment 2200 in which some systems and methods of the present invention may be implemented. As an example, the process controller 108 and the computing device 110 (or a component thereof) in Fig. 1 may represent a computer 2201, as shown in Fig. 22. The computer 2201 may include one or more components, such as one or more processors 2203, a system memory 2212, and a bus 2213 that connects various components of the computer 2201, including one or more processors 2203, to the system memory 2212. In the case of multiple processors 2203, the system may use parallel computing.

The bus 2213 may include one or more of several possible types of bus structures, such as a memory bus, a memory controller, a peripheral bus, an accelerated graphics port, and a processor or local bus using any of a variety of bus structures. The bus 2213 and all buses specified in this description may also be implemented via a wired or wireless network connection.

Computer 2201 typically includes various computer-readable media. An exemplary computer-readable media may be any available media that is accessible to computer 2201 and includes, for example and not intended to be limiting, both volatile and nonvolatile media, removable and non-removable media. System memory 2212 may include computer-readable media in the form of volatile memory, such as random access memory (RAM), and/or nonvolatile memory, such as read-only memory (ROM). The system memory 2212 may typically contain data, such as chromatography data 2207 and/or program modules, such as an operating system 2205 and chromatography software 2206, which are accessed and/or controlled by one or more processors 2203. Numerous features and advantages of the present disclosure are apparent from the detailed description, and thus the appended claims are intended to cover all such features and advantages of the present invention that fall within the true spirit and scope of the invention. Furthermore, since numerous modifications and variations may easily occur to those skilled in the art, it is not desirable to limit the present invention to the precise construction and function illustrated and described, and accordingly, all suitable modifications and equivalents may be resorted to without departing from the scope of the present invention.

In another aspect, the computer 2201 may also include other removable/non-removable, volatile/non-volatile computer storage media. The storage device 2204 may provide non-volatile storage of computer code, machine-readable instructions, data structures, program modules and other data for the computer 2201. For example, the storage device 2204 may represent a hard disk, a removable magnetic disk, a removable optical disk, magnetic cassettes or other magnetic storage devices, flash memory cards, CD-ROMs, digital versatile disks (DVDs) or other optical storage, random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM) and the like.

It will be understood by those skilled in the art that the concept on which the present invention is based can be easily used as a basis for developing other structures, methods and systems for carrying out several objects of the present disclosure. Accordingly, the claims should not be considered as limited by the preceding description.

-

Claims (7)

CLAUSE OF THE INVENTION 1. A method for monitoring a chromatographic process performed using a chromatographic column having an input and an output, comprising obtaining raw chromatographic data collected during the performance of the chromatographic process from a plurality of signals provided by one or more detectors, wherein each signal of the plurality of signals is associated with one of a plurality of units, and each unit corresponds to a stage in the chromatographic process; obtaining a subset of data from said raw chromatographic data corresponding to at least one of the signals of one or more detectors that is suitable for performing a transition analysis, wherein the transition is a response at the column outlet to a change in stage at the column inlet when one mobile phase is replaced by another; wherein obtaining a subset of said data involves using a profile that provides instructions for applying one or more criteria for selecting raw chromatographic data, wherein the profile is formed on the basis of previous chromatographic analyses; performing a transition analysis on said subset of data, wherein performing the transition analysis includes generating a function using said subset of data; and analyzing said function to generate performance parameters; and performing an action based on the performance parameters, wherein performing the action includes at least one of the following: generating an event notification; generating an event rating; or generating a deviation notification with data characterizing the deviation. 2. The method according to claim 1, further comprising selecting a combination of the first block and the first signal in accordance with a profile defining a plurality of selection criteria. 3. The method according to claim 2, wherein the plurality of selection criteria comprise: to what extent blocks appear at regular intervals in the chromatographic cycle; the degree to which one of a plurality of signals saturates a detector; the degree to which a set of signals approaches a stationary phase at a detectable level; the amount of variation in a set of signals; or the number of inflection points exhibited by a set of signals during a transition phase. 4. The method according to claim 1, wherein performing transition analysis on said subset of data includes performing transition analysis on processed chromatographic data, and the method further comprises generating processed chromatographic data by applying a noise reduction method to said subset of data. 5. The method according to claim 1, wherein the noise reduction method comprises selecting a portion of said subset of data for analysis using predetermined set values; normalization of the specified part to prevent deviation of the value; applying at least one smoothing filter on said portion to generate smoothed data; analysis of the specified part for dynamic signal errors. 6. The method according to claim 5, wherein the noise reduction method further comprises selecting smoothed data corresponding to a chromatographic transition feature, wherein the feature includes one of the duration of the derivative; maximum intensity; duration from initiation; or expected background noise of the sensor. 7. A method for monitoring a chromatographic process performed using a chromatographic column having an input and an output, comprising obtaining raw chromatographic data collected during the performance of the chromatographic process from a plurality of signals provided by one or more detectors, wherein each signal of the plurality of signals is associated with one of a plurality of units, and each unit corresponds to a stage in the chromatographic process; obtaining a subset of data from said raw chromatographic data corresponding to at least one of the signals of one or more detectors that is suitable for performing a transition analysis, wherein the transition is a response at the column outlet to a change in stage at the column inlet when one mobile phase is replaced by another; performing a transition analysis on a specified subset of data, wherein performing the transition analysis includes: -