CN114242169A - Antigen epitope prediction method for B cells - Google Patents
Antigen epitope prediction method for B cells Download PDFInfo
- Publication number
- CN114242169A CN114242169A CN202111537519.XA CN202111537519A CN114242169A CN 114242169 A CN114242169 A CN 114242169A CN 202111537519 A CN202111537519 A CN 202111537519A CN 114242169 A CN114242169 A CN 114242169A
- Authority
- CN
- China
- Prior art keywords
- amino acid
- action
- state
- sequence
- value
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 49
- 210000003719 b-lymphocyte Anatomy 0.000 title claims abstract description 32
- 239000000427 antigen Substances 0.000 title description 3
- 102000036639 antigens Human genes 0.000 title description 3
- 108091007433 antigens Proteins 0.000 title description 3
- 230000009471 action Effects 0.000 claims abstract description 112
- 239000003795 chemical substances by application Substances 0.000 claims abstract description 35
- 230000006870 function Effects 0.000 claims abstract description 33
- 125000003275 alpha amino acid group Chemical group 0.000 claims abstract description 28
- 102000004169 proteins and genes Human genes 0.000 claims abstract description 27
- 108090000623 proteins and genes Proteins 0.000 claims abstract description 27
- 238000012549 training Methods 0.000 claims abstract description 23
- 230000000295 complement effect Effects 0.000 claims abstract description 15
- 125000000539 amino acid group Chemical group 0.000 claims abstract description 12
- 230000008859 change Effects 0.000 claims abstract description 8
- 150000001413 amino acids Chemical class 0.000 claims description 52
- 238000004364 calculation method Methods 0.000 claims description 10
- 230000008569 process Effects 0.000 claims description 7
- 210000004027 cell Anatomy 0.000 claims description 2
- 238000000605 extraction Methods 0.000 claims 1
- 238000013461 design Methods 0.000 description 3
- 230000003993 interaction Effects 0.000 description 3
- 230000007246 mechanism Effects 0.000 description 3
- 230000001186 cumulative effect Effects 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 238000011156 evaluation Methods 0.000 description 2
- 230000007774 longterm Effects 0.000 description 2
- 238000010801 machine learning Methods 0.000 description 2
- 208000035473 Communicable disease Diseases 0.000 description 1
- 230000007815 allergy Effects 0.000 description 1
- 230000000975 bioactive effect Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000003111 delayed effect Effects 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 201000010099 disease Diseases 0.000 description 1
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 description 1
- 229940079593 drug Drugs 0.000 description 1
- 239000003814 drug Substances 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 239000000284 extract Substances 0.000 description 1
- 239000012634 fragment Substances 0.000 description 1
- 238000009169 immunotherapy Methods 0.000 description 1
- 230000010354 integration Effects 0.000 description 1
- 244000005700 microbiome Species 0.000 description 1
- 238000000455 protein structure prediction Methods 0.000 description 1
- 230000002787 reinforcement Effects 0.000 description 1
- 238000012827 research and development Methods 0.000 description 1
- 230000007704 transition Effects 0.000 description 1
- 229960005486 vaccine Drugs 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16B—BIOINFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR GENETIC OR PROTEIN-RELATED DATA PROCESSING IN COMPUTATIONAL MOLECULAR BIOLOGY
- G16B30/00—ICT specially adapted for sequence analysis involving nucleotides or amino acids
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16B—BIOINFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR GENETIC OR PROTEIN-RELATED DATA PROCESSING IN COMPUTATIONAL MOLECULAR BIOLOGY
- G16B40/00—ICT specially adapted for biostatistics; ICT specially adapted for bioinformatics-related machine learning or data mining, e.g. knowledge discovery or pattern finding
Landscapes
- Life Sciences & Earth Sciences (AREA)
- Physics & Mathematics (AREA)
- Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Engineering & Computer Science (AREA)
- General Health & Medical Sciences (AREA)
- Spectroscopy & Molecular Physics (AREA)
- Biophysics (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Biotechnology (AREA)
- Evolutionary Biology (AREA)
- Theoretical Computer Science (AREA)
- Artificial Intelligence (AREA)
- Chemical & Material Sciences (AREA)
- Analytical Chemistry (AREA)
- Proteomics, Peptides & Aminoacids (AREA)
- Bioethics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Data Mining & Analysis (AREA)
- Databases & Information Systems (AREA)
- Epidemiology (AREA)
- Evolutionary Computation (AREA)
- Public Health (AREA)
- Software Systems (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
A method for epitope prediction of B cells, said method comprising first forming a pre-training set PT; in each epsilon of the Q _ learning algorithm, the Q-agent takes any 8 consecutive amino acid residues in the primary sequence of the protein as states, and selects k residues from the 12 consecutive residues following each state to incorporate into the state as a first action; selecting one of n complementary classifiers as a second action option, searching in PT according to a continuous action search method, giving instant reward to the searched amino acid sequence by a tendency reward rule, calculating a Q value and updating until the change of the cost function is less than 1 percent, and finishing training; then, the amino acid sequence is searched in the primary protein sequence by using the trained strategy, and the primary protein sequence is classified by the selected classifier. The invention greatly enhances the prediction capability of the B cell epitope through automatic iteration and improves the accuracy of epitope classification.
Description
Technical Field
The invention relates to a B cell epitope prediction method, which can accurately predict B cell epitopes and belongs to the technical field of artificial intelligent detection of microorganisms.
Background
The accurate determination of B cell antigen epitope is an important basis for the design of bioactive drugs and epitope vaccines, is a key step in the research and development of disease kits, and is a basic technology for researching immunodiagnosis and immunotherapy. Machine learning-based B cell epitope prediction is an important technical route for determining the epitope, and has the advantages of greatly saving time, money and labor cost compared with other technical routes.
SEPPA is an epitope prediction software recommended in an Immune Epitope Database (IEDB) built by national institute of allergy and infectious diseases, and is updated to 3.0 edition in 2019. The scholars responsible for the development of SEPPA 3.0 have noted in their paper that the progress of conformational epitope prediction has progressed smoothly but slowly over the last decade.
The existing epitope prediction adopts a supervised learning strategy, and a class predictor is obtained by learning epitope samples and non-epitope samples. Although new epitope prediction methods are continuously generated, the prediction accuracy is continuously improved, and the problems of low universality of the model, low classification accuracy, slow updating of the prediction model and the like exist. In particular, the number of amino acids in the prediction result is preset as an integer before prediction by a common window method, which is highly artificial and difficult to predict an epitope with an optimal length.
With the revolutionary breakthrough of AlphaFold in the field of protein structure prediction as AlphaGo defeats the strongest human player in go wars, these successes have greatly inspired us. The two breakthroughs have a common characteristic that an automatic learning mechanism is introduced, so that the model can continuously self-iterate to gradually generate strong recognition capability.
However, the existing methods are non-automatic learning and cannot enhance the prediction capability through automatic iteration, so that an automatic mechanism is introduced into the prediction of the epitope of the B cell, and a method capable of accurately determining the epitope of the B cell is necessary to design.
Disclosure of Invention
The invention aims to provide a B cell epitope prediction method aiming at the defects of the prior art so as to improve the accuracy of B cell epitope prediction.
The problems of the invention are solved by the following technical scheme:
a method for predicting B cell epitope, the method searches B cell epitope sequence data from IEDB database to form set EPT, extracts corresponding protein primary sequence from uniport database to form pre-training set PT; based on a Q _ learning algorithm, changing one action of the algorithm into two actions for training; in each epsilon, the Q-agent takes any 8 consecutive amino acid residues in the primary sequence of the protein as states, and selects k residues from the last 12 consecutive residues of each state to incorporate into that state as a first action; selecting one of n complementary classifiers as a second action option, searching in a primary protein sequence in PT according to a continuous action search method, giving instant reward to the searched amino acid sequence by a tendency reward rule, calculating a Q value and updating until the change of a value function is less than 1 percent, and finishing training; then, the amino acid sequence is searched in the primary protein sequence by using the trained strategy, and the selected classifier is used for classifying, so that the B cell epitope prediction is realized.
The method for predicting the epitope of the B cell comprises the following steps:
a. b cell epitope sequence data are searched from an IEDB database to form a set EPT, corresponding protein primary sequences are extracted from a uniport database to form a pre-training set PT, and a set containing n more than or equal to 2 complementary classifiers is selected as a second action;
b. taking any 8 continuous amino acid residues in the primary sequence of the protein as states, and selecting k residues from 12 continuous residues after each state to be incorporated into the state as a first action; selecting one of n complementary classifiers as a second action option, initializing Q values corresponding to all the states and actions to 0, setting the learning rate alpha to be an arbitrary number between 0 and 1, setting the discount factor gamma to be an arbitrary number between 0 and 1, setting the value of the epamode, and setting the initial state s0Is any 8 amino acid residues of the pre-training set;
c. in each epsilon, the Q-agent performs a search in the primary sequence of the proteins in the set PT according to a continuous motion search method: in the tth step, the Q agent selects an action from the first set of actionsMaking
Then selecting an action in the set of second actions
Awarding a reward R according to a tendency reward law after the two actions have been performedtAnd the next observation state st+1Then updating the Q value, updating the state and the action table at the same time, and ending the search training process when the change of the value function is less than 1%;
d. and searching out amino acid combinations in each protein primary sequence by utilizing a strategy obtained by training, classifying by using a selected classifier, and if the result of the classifier shows that the searched amino acid sequences are epitopes, considering the amino acid sequences as B cell epitopes, otherwise, not considering the amino acid sequences as B cell epitopes.
The epitope prediction method for B cells described above, the specific search process of the continuous motion search method is:
taking any 8 amino acid residues of each protein primary sequence as initial state s0And the corresponding amino acid sequence is marked as X1X2…X8, wherein XjDenotes the j-th amino acid, j ═ 1,2, …,8, in order to change from the initial state s0Selecting k residues from the next 12 consecutive residues to incorporate into the state as a first action, 1 ≦ k ≦ 12, to select one of the n complementary classifiers as a second action option; according to the corresponding
Value selection first action and second action, where a1,a2All possible actions in the first action and all possible actions in the second action are referred to respectively, then the reward is calculated by the tendency reward rule for the above two actions, and the cost function is calculated according to the following formula:
wherein ,Vπ(s) is the cost function at state s, π is the strategy,
is that R istIs the yield of step t, V(s)t+1) Is the next state st+1A cost function of;
the Q value was calculated as follows:
wherein Qπ(s,a1,a2) Is a cost function that performs two consecutive actions in state s,
is the next state st+1Perform two consecutive actions
A cost function of (a);
the Q value is also updated as follows:
then changing the state, repeating the above steps, and updating the Q value.
The epitope prediction method for B cells comprises the following steps:
extracting the characteristics of the amino acid sequence searched by the first action as the input of a classifier selected by the second action, and calculating the classification score SC of the amino acid sequence by the classifiertIn the collective EPT, the probability of occurrence of each amino acid and the probability of occurrence of an amino acid pair comprising two consecutive amino acids are calculated, and any one of the amino acids as isiProbability of occurrence
The calculation was performed as follows:
wherein asiRepresents any one of the 20 amino acids, num (as)i) Represents asiThe number of times the EPT set occurs, maxnum (as)1,as2,…,as20) Indicates the maximum number of occurrences of 20 amino acids in the pool EPT, minnum (as)1,as2,…,as20) Represents the minimum number of occurrences of 20 amino acids in the set EPT;
any one amino acid pair AAiProbability of occurrence of
The calculation was performed as follows:
wherein AAiRepresents one of 400 amino acid pairs, num (AA)i) Represents AAiThe number of occurrences of the set EPT, maxnum (AA)1,AA2,…,AA400) Represents the maximum number of occurrences of 400 amino acid pairs in the EPT pool, minnum (AA)1,AA2,…,AA400) Represents the minimum number of occurrences of 400 amino acids in the set EPT;
the timely reward for the amino acid sequence generated in step t is calculated by:
wherein len (sq)t) Representing the sequence sqtContains the number of amino acids, len (sq)t) Indicates the number of consecutive amino acid pairs.
Advantageous effects
The method combines the Q _ learning algorithm, the continuous action search method and the tendency reward method together, and simultaneously introduces the complementary classifier, so that the method greatly enhances the prediction capability of the B cell epitope through automatic iteration and improves the accuracy of epitope classification.
Drawings
The present invention will be described in further detail with reference to the accompanying drawings.
FIG. 1 is a diagram of the Q _ learning algorithm.
Detailed Description
The invention provides an epitope prediction method for B cells, which adopts an epitope prediction method of continuous action search based on a Q-learning method, one action selects sequence length, and one action selects a complementary classifier, thereby realizing the independent selection of sequence length and the selection of the optimal classifier for classification.
In the Q-Learning algorithm, each state-action has a corresponding Q value. Therefore, the Learning process of the Q-Learning algorithm is to iteratively calculate the Q value of the Learning state-action pair. And finally, the optimal action strategy obtained by the learner is the action corresponding to the maximum Q value in the selected state s. The Q value Q (s, a) based on action a in state s is defined as the cumulative reward value obtained after the learner executes action a in state s and then executes it according to a certain action strategy. The basic equation for updating the Q value is as follows:
Q(st,at)＝Q(st,at)+α[Rst+γmaxQ(st+1，a)-Q(st，at)]
in the above formula (wherein, a is an optional action in the state; Rs)tAn immediate reward given to the environment at time t state s; α is the learning rate; q(s)t,at) At time t, the evaluation value of the state-action (s, a).
The Q-Learning algorithm pseudo code is shown in Table 1:
TABLE 1Q-Learning Algorithm pseudocode
FIG. 1 is a diagram of the Q _ learning algorithm.
Interpretation of professional terms
(1) The primary sequence of the protein is a sequence consisting of 20 amino acids, such as ADFCEGHIKLST.
(2) A B cell epitope, which is a portion of the primary sequence of a protein, may consist of a partial sequence.
(3) "epamode" is the process by which an Agent (Agent) executes a policy within an environment from start to finish.
(4) Agent (Agent): a software and hardware mechanism. It takes corresponding measures through interaction with the surrounding environment. Agent chinese is also called Agent, often called Q Agent in Q _ learning algorithm, which is the subject to exploration and learning in the environment.
(5) Action (Action): various possible actions that the agent may take. Although the actions themselves are somewhat self-explanatory, we still need to have agents that can choose from a series of discrete, and possible actions.
(6) Environment (Environment): there is an interaction between the external environment and the agent, and a responsive relationship. The environment takes as input the current state and actions of the agent and as output the reward and next state of the agent. The context is all present except the agent.
(7) State (State): the state is a specific and direct case that the agent finds itself, including: a particular location, time of day, and an instantaneous configuration that associates the agent with other important things.
(8) Reward (Reward): rewards are feedback by which we can measure the success or failure of various actions of an agent in a given state.
(9) Discount factor (Discount factor): the discount factor is a multiplier. The future rewards discovered by the agent are multiplied by the factor to attenuate the cumulative impact of such rewards on the agent's current action selection. This is the core of reinforcement learning, namely by gradually reducing the value of future rewards in order to give more weight to recent actions. This is crucial for the paradigm based on the principle of "delayed action".
(10) Policy (Policy): is a function of the input state observation and the output action. Is the policy that the agent uses to determine the next action based on the current state. It can map different states to various actions to promise the highest reward.
(11) Value (Value): it is defined as a long-term expected reward (not a short-term reward) with a discount on the current state under a particular policy. A short-term reward is an instant reward that an agent gets in a certain state and takes a certain action. The value is the total amount of rewards an agent expects to receive from a certain state until the future.
(12) Q value (Q-value) or action-value (action-value): the difference from "value" is that the Q value requires an additional parameter, i.e. the current action. It refers to a long-term reward that an action generates from the current state under a particular policy.
(13) Bellman equation: it is a set of equations that decompose the value function into an instant prize plus a discount future value.
(14) Iteration of values (Value iteration): this is an algorithm that iteratively refines the estimate of the value to compute a function with the best state value. The algorithm initializes the value function to an arbitrary random value and then repeatedly updates the values of the Q value and the value function until they converge.
(15) Policy iteration (Policy iteration): since the agent is only interested in finding the optimal strategy, the optimal strategy will sometimes converge before the cost function. Thus, a policy iteration should not repeatedly improve the evaluation of the value function, but rather, need to redefine the policy at each step and compute values from the new policy until the policy converges.
(16) Q learning (Q-learning): as an example of a model-less learning algorithm, it does not assume that the agent has mastered the transition to state and the reward model, but rather "thinks" that the agent will find the correct action through trial and error. Therefore, the basic idea of Q learning is: during agent and environment interaction, the Q function of the state-action pairs is approached by observing the sample of the Q value function.
Most of epitope prediction methods based on machine learning adopt a marked sequence as a positive sample and an unmarked sequence as a negative sample, utilize amino acid physicochemical characteristics, statistical properties, structural information and the like as characteristic input, train a classifier by utilizing a common classification learning algorithm, and then classify the sequence by using the classifier. This prediction method generally sets a window in advance, and the window size is how many amino acids the result contains. Because the epitope sequence has larger difference, the epitope can not be predicted well by using a trained classifier, so the integration method has more advantages.
Firstly, searching B cell epitope sequence data from an IEDB database to form a set EPT, extracting a corresponding protein primary sequence from a uniport database to form a pre-training set PT; in each epsilon, the Q-agent searches for a combination of residues in the PT according to a continuous action search method, taking any 8 consecutive amino acid residues in the PT as a state, and selecting k residues from the 12 consecutive residues following each state to incorporate into the state as a first action; selecting one of n complementary classifiers as a second action option, giving instant reward to the searched amino acid sequence by a tendency reward rule, calculating a Q value, updating, ending training until the change of a cost function is less than 1%, and then predicting the epitope of the B cell by using the trained classifier.
The specific search steps are as follows:
firstly, B cell epitope sequence data are searched from an IEDB database to form a set EPT, a corresponding protein primary sequence is extracted from a uniport database to form a pre-training set PT, and a set containing n ≧ 2 complementary classifiers is selected as a second action.
Secondly, initializing Q values corresponding to all states and actions to be 0, setting the learning rate alpha to be an arbitrary number between 0 and 1, setting the discount factor gamma to be an arbitrary number between 0 and 1, setting the numerical value of the epamode, and setting the state s0Is an arbitrary of a pre-training setMeaning 8 amino acid residues.
Third, in each epsilon, the Q agent searches in the set PT according to a continuous action search method, and in the t step, the Q agent selects an action from the first set of actions
Then selecting an action in the set of second actions
The two actions are performed continuously, and after the two actions are performed, a reward R is given according to a tendency reward ruletAnd the next observation state st+1. The Q value is then updated, along with the state and action tables. When the variance of the cost function is less than 1%, the search training process ends.
And fourthly, searching out amino acid combinations in each protein primary sequence by utilizing a strategy obtained by training, classifying by a selected classifier, and considering the searched amino acid sequences as epitopes if the results of the classifier show that the amino acid sequences are the epitopes, or not considering the amino acid sequences as the epitopes.
The "continuous search method" in the third step is a method in which any 8 amino acid residues in the primary sequence of each protein are used as the initial state s0And the corresponding sequence is marked as X1X2…X8By s0The latter i residues are combined into action, i ═ 1,2, …, 12. According to the corresponding
Value and state s0Selecting a first action and a second action, wherein a1,a2Respectively, all possible actions in the first action and all possible actions in the second action, performing the action
Corresponding to the initial state s0The first action, which is the selection of k residues from the next 12 consecutive residues to incorporate into this state, results in a first amino acid sequence of length k +8Column fragment X1X2…X8…Xk+8Performing an action
Selecting the mth classifier in the second action set, wherein m is more than or equal to 1 and less than or equal to n, calculating the reward for the two actions by the tendency reward rule, calculating the cost function according to the formula (1), calculating the Q value according to the formula (2), and observing the next state s1And meanwhile, updating the Q value according to the formula (3). Then, in
And state s1Selection actions
And
calculating a cost function according to formula (1), calculating a Q value according to formula (2), and observing a next state s2And meanwhile, updating the Q value according to the formula (3). Thereafter, training is performed in the same manner as described above.
The value function under each set of successive actions in the learning network is:
wherein ,Vπ(s) is the cost function at state s, π is the strategy,
is that R istIs the yield of step t, V(s)t+1) Is the next state st+1The following cost function.
The Q value calculation formula of each group of actions under the step t is as follows:
wherein Qπ(s,a1,a2) Is a cost function that performs two consecutive actions in state s,
is the next state st+1Perform two consecutive actions
A cost function of (2).
The updating calculation method of the Q value comprises the following steps:
the tendency reward rule in the third step is calculated by the classification score given to the sequence by the classifier and the occurrence probability of the sequence. Extracting the features of the amino acid sequence searched for in the first action as the input of the classifier selected in the second action, calculating the classification score of the amino acid sequence by the classifier, for example, in step t, the score calculated by the classifier selected in the second action for the amino acid sequence obtained in the first action is recorded as SCt. In the set EPT, the number of occurrences of each amino acid and the occurrence of pairs of amino acids comprising two consecutive amino acids are counted. For any one amino acid asiProbability of occurrence
According to the formula
Calculation of, wherein asiRepresents any one of the 20 amino acids, num (as)i) Represents asiThe number of times the EPT set occurs, maxnum (as)1,as2,…,as20) Indicates the maximum number of occurrences of 20 amino acids in the pool EPT, minnum (as)1,as2,…,as20) Represents the minimum number of occurrences of 20 amino acids in the set EPT. Any one amino acid pairAAiProbability of occurrence of
According to the formula
Performing a calculation in which AAiRepresents one of 400 amino acid pairs, num (AA)i) Represents AAiThe number of occurrences of the set EPT, maxnum (AA)1,AA2,…,AA400) Represents the maximum number of occurrences of 400 amino acid pairs in the EPT pool, minnum (AA)1,AA2,…,AA400) Represents the minimum number of occurrences of 400 amino acids in the set EPT.
The timely reward obtained by the amino acid sequence generated in step t is given by the formula
Calculation of len (sq) thereint) Representing the sequence sqtContains the number of amino acids, len (sq)t) Indicates the number of consecutive amino acid pairs.
Compared with the traditional epitope prediction scheme based on a window method, the epitope prediction method based on continuous action search realizes autonomous selection of the epitope sequence, eliminates the influence of human factors, and improves the classification accuracy by introducing a complementary classifier.
The invention adopts a 'continuous action search method', which takes the number of amino acids in a common epitope sequence of 8-20 as a design basis and meets the calculation requirement of a complementary classifier under the condition of covering the length of the common epitope.
The invention adopts a tendency reward rule, takes the calculated score of the classifier on the search sequence and the statistical characteristics of the sequence into consideration, comprehensively judges the score of the classifier, the occurrence probability of the amino acid and the occurrence probability of the amino acid pair, and calculates the timely reward of two continuous actions in each state.
Claims (4)
1. A method for predicting B cell epitope is characterized in that firstly B cell epitope sequence data are searched from an IEDB database to form a set EPT, and corresponding protein primary sequences are extracted from a uniport database to form a pre-training set PT; based on a Q _ learning algorithm, changing one action of the algorithm into two actions for training; in each epsilon, the Q-agent takes any 8 consecutive amino acid residues in the primary sequence of the protein as states, and selects k residues from the last 12 consecutive residues of each state to incorporate into that state as a first action; selecting one of n complementary classifiers as a second action option, searching in a primary protein sequence in PT according to a continuous action search method, giving instant reward to the searched amino acid sequence by a tendency reward rule, calculating a Q value and updating until the change of a value function is less than 1 percent, and finishing training; then, the amino acid sequence is searched in the primary protein sequence by using the trained strategy, and the selected classifier is used for classifying, so that the B cell epitope prediction is realized.
2. The method of predicting an epitope for a B cell according to claim 1, comprising the steps of:
a. b cell epitope sequence data are searched from an IEDB database to form a set EPT, corresponding protein primary sequences are extracted from a uniport database to form a pre-training set PT, and a set containing n more than or equal to 2 complementary classifiers is selected as a second action;
b. taking any 8 continuous amino acid residues in the primary sequence of the protein as states, and selecting k residues from 12 continuous residues after each state to be incorporated into the state as a first action; selecting one of n complementary classifiers as a second action option, initializing Q values corresponding to all the states and actions to 0, setting the learning rate alpha to be an arbitrary number between 0 and 1, setting the discount factor gamma to be an arbitrary number between 0 and 1, setting the value of the epamode, and setting the initial state s0Is any 8 amino acid residues of the pre-training set; c. in each epsilon, the Q-agent performs the proteins in the set PT according to a continuous motion search methodSearch in the primary sequence: in the tth step, the Q-agent selects an action from the first set of actions
Then selecting an action in the set of second actions
Awarding a reward R according to a tendency reward law after the two actions have been performedtAnd the next observation state st+1Then updating the Q value, updating the state and the action table at the same time, and ending the search training process when the change of the value function is less than 1%;
d. and searching out amino acid combinations in each protein primary sequence by utilizing a strategy obtained by training, classifying by using a selected classifier, and if the result of the classifier shows that the searched amino acid sequences are epitopes, considering the amino acid sequences as B cell epitopes, otherwise, not considering the amino acid sequences as B cell epitopes.
3. The method of claim 2, wherein the specific search process of the continuous motion search method is:
taking any 8 amino acid residues of each protein primary sequence as initial state s0And the corresponding amino acid sequence is marked as X1X2…X8, wherein XjDenotes the j-th amino acid, j ═ 1,2, …,8, in order to change from the initial state s0Selecting k residues from the next 12 consecutive residues to incorporate into the state as a first action, 1 ≦ k ≦ 12, to select one of the n complementary classifiers as a second action option; according to the corresponding
Value selection first action and second action, where a1,a2All possible actions in the first action and all possible actions in the second action are referred to separately, and then the two actions are counted by the tendency reward ruleCalculating the reward and calculating the cost function according to the following formula:
wherein ,Vπ(s) is the cost function at state s, π is the strategy,
is that R istIs the yield of step t, V(s)t+1) Is the next state st+1A cost function of;
the Q value was calculated as follows:
wherein Qπ(s,a1,a2) Is a cost function that performs two consecutive actions in state s,
is the next state st+1Perform two consecutive actions
A cost function of (a);
the Q value is also updated as follows:
then changing the state, repeating the above steps, and updating the Q value.
4. The method of claim 3, wherein the trend reward algorithm is:
extraction of features from the amino acid sequence searched for in the first actionThe input of the classifier selected for the second action, the classifier calculates the classification score SC of the amino acid sequencetIn the collective EPT, the probability of occurrence of each amino acid and the probability of occurrence of an amino acid pair comprising two consecutive amino acids are calculated, and any one of the amino acids as isiProbability of occurrence
The calculation was performed as follows:
wherein asiRepresents any one of the 20 amino acids, num (as)i) Represents asiThe number of times the EPT set occurs, maxnum (as)1,as2,…,as20) Indicates the maximum number of occurrences of 20 amino acids in the pool EPT, minnum (as)1,as2,…,as20) Represents the minimum number of occurrences of 20 amino acids in the set EPT;
any one amino acid pair AAiProbability of occurrence of
The calculation was performed as follows:
wherein AAiRepresents one of 400 amino acid pairs, num (AA)i) Represents AAiThe number of occurrences of the set EPT, maxnum (AA)1,AA2,…,AA400) Represents the maximum number of occurrences of 400 amino acid pairs in the EPT pool, minnum (AA)1,AA2,…,AA400) Represents the minimum number of occurrences of 400 amino acids in the set EPT;
the timely reward for the amino acid sequence generated in step t is calculated by:
wherein len (sq)t) Representing the sequence sqtContains the number of amino acids, len (sq)t) Indicates the number of consecutive amino acid pairs.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111537519.XA CN114242169B (en) | 2021-12-15 | 2021-12-15 | Antigen epitope prediction method for B cells |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111537519.XA CN114242169B (en) | 2021-12-15 | 2021-12-15 | Antigen epitope prediction method for B cells |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114242169A true CN114242169A (en) | 2022-03-25 |
| CN114242169B CN114242169B (en) | 2023-10-20 |
Family
ID=80756774
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111537519.XA Active CN114242169B (en) | 2021-12-15 | 2021-12-15 | Antigen epitope prediction method for B cells |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114242169B (en) |
Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20100279881A1 (en) * | 2007-02-02 | 2010-11-04 | Sompuram Seshi R | Epitope-mediated antigen prediction |
| WO2013177214A2 (en) * | 2012-05-21 | 2013-11-28 | Distributed Bio Inc | Epitope focusing by variable effective antigen surface concentration |
| CN105868583A (en) * | 2016-04-06 | 2016-08-17 | 东北师范大学 | Method for predicting epitope through cost-sensitive integrating and clustering on basis of sequence |
| CN107033226A (en) * | 2017-06-27 | 2017-08-11 | 中国农业科学院兰州兽医研究所 | A kind of PPR virus F protein epitope peptide and its determination, preparation method and application |
| WO2017184590A1 (en) * | 2016-04-18 | 2017-10-26 | The Broad Institute Inc. | Improved hla epitope prediction |
| CN107341363A (en) * | 2017-06-29 | 2017-11-10 | 河北省科学院应用数学研究所 | A kind of Forecasting Methodology of proteantigen epitope |
| CN107909153A (en) * | 2017-11-24 | 2018-04-13 | 天津科技大学 | The modelling decision search learning method of confrontation network is generated based on condition |
| JP6500144B1 (en) * | 2018-03-28 | 2019-04-10 | Ｋｏｔａｉバイオテクノロジーズ株式会社 | Efficient clustering of immune entities |
| US20200143206A1 (en) * | 2018-11-05 | 2020-05-07 | Royal Bank Of Canada | System and method for deep reinforcement learning |
| CN112106141A (en) * | 2018-03-16 | 2020-12-18 | 弘泰生物科技股份有限公司 | Efficient clustering of immune entities |
-
2021
- 2021-12-15 CN CN202111537519.XA patent/CN114242169B/en active Active
Patent Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20100279881A1 (en) * | 2007-02-02 | 2010-11-04 | Sompuram Seshi R | Epitope-mediated antigen prediction |
| WO2013177214A2 (en) * | 2012-05-21 | 2013-11-28 | Distributed Bio Inc | Epitope focusing by variable effective antigen surface concentration |
| CN105868583A (en) * | 2016-04-06 | 2016-08-17 | 东北师范大学 | Method for predicting epitope through cost-sensitive integrating and clustering on basis of sequence |
| WO2017184590A1 (en) * | 2016-04-18 | 2017-10-26 | The Broad Institute Inc. | Improved hla epitope prediction |
| CN107033226A (en) * | 2017-06-27 | 2017-08-11 | 中国农业科学院兰州兽医研究所 | A kind of PPR virus F protein epitope peptide and its determination, preparation method and application |
| CN107341363A (en) * | 2017-06-29 | 2017-11-10 | 河北省科学院应用数学研究所 | A kind of Forecasting Methodology of proteantigen epitope |
| CN107909153A (en) * | 2017-11-24 | 2018-04-13 | 天津科技大学 | The modelling decision search learning method of confrontation network is generated based on condition |
| CN112106141A (en) * | 2018-03-16 | 2020-12-18 | 弘泰生物科技股份有限公司 | Efficient clustering of immune entities |
| JP6500144B1 (en) * | 2018-03-28 | 2019-04-10 | Ｋｏｔａｉバイオテクノロジーズ株式会社 | Efficient clustering of immune entities |
| US20200143206A1 (en) * | 2018-11-05 | 2020-05-07 | Royal Bank Of Canada | System and method for deep reinforcement learning |
Non-Patent Citations (1)
| Title |
|---|
| 马骋乾;谢伟;孙伟杰;: "强化学习研究综述", 指挥控制与仿真, no. 06 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114242169B (en) | 2023-10-20 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Li et al. | Self-paced ARIMA for robust time series prediction | |
| CN114841257B (en) | A small sample target detection method based on self-supervised contrast constraints | |
| CN111861909B (en) | Network fine granularity image classification method | |
| CN111310799B (en) | Active learning method based on historical evaluation result | |
| CN112085559A (en) | Interpretable product recommendation method and system based on temporal knowledge graph | |
| CN113283426A (en) | Embedded target detection model generation method based on multi-target neural network search | |
| CN109215740A (en) | Full-length genome RNA secondary structure prediction method based on Xgboost | |
| CN108596327A (en) | A kind of seismic velocity spectrum artificial intelligence pick-up method based on deep learning | |
| CN116153495A (en) | A method for predicting the prognosis and survival of patients with esophageal cancer following immunotherapy | |
| US20250036969A1 (en) | Multilingual event causality identification method and system based on meta-learning with knowledge | |
| CN115359292B (en) | Standing long jump stage classification method based on feature adaptive fusion | |
| CN116226629B (en) | Multi-model feature selection method and system based on feature contribution | |
| CN114169595B (en) | Emotion prediction method based on big data | |
| CN114708903A (en) | Method for predicting distance between protein residues based on self-attention mechanism | |
| CN109558898A (en) | A kind of more options learning method of the high confidence level based on deep neural network | |
| CN107945534A (en) | A kind of special bus method for predicting based on GMDH neutral nets | |
| CN111144462B (en) | Unknown individual identification method and device for radar signals | |
| CN112651499A (en) | Structural model pruning method based on ant colony optimization algorithm and interlayer information | |
| Shao et al. | Visual navigation with actor-critic deep reinforcement learning | |
| CN116595371A (en) | Topic heat prediction model training method, topic heat prediction method and topic heat prediction device | |
| CN115512272A (en) | Time sequence event detection method for multi-event instance video | |
| CN109378034B (en) | A protein prediction method based on distance distribution estimation | |
| CN110334080A (en) | A kind of construction of knowledge base method for realizing autonomous learning | |
| CN114242169A (en) | Antigen epitope prediction method for B cells | |
| CN113035363B (en) | Probability density weighted genetic metabolic disease screening data mixed sampling method |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |