[go: up one dir, main page]
Scribd
Upload
95 views17 pages

Code Oean - GPU Accelerated Workflows Webinar Slides

This document discusses how GPU acceleration with NVIDIA and Code Ocean can help accelerate single-cell genomics workflows and analysis. It highlights how GPUs can speed up processing by around 20x compared to CPUs. The webinar demonstrates how Code Ocean compute capsules containing RAPIDS and ATACworks toolkits can provide a flexible computational environment for researchers to perform GPU-accelerated single-cell analysis on large datasets in a reproducible manner. This helps address challenges of growing dataset sizes and complexity for computational researchers.

Uploaded by

Namrata Lakhani
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
95 views17 pages

Code Oean - GPU Accelerated Workflows Webinar Slides

This document discusses how GPU acceleration with NVIDIA and Code Ocean can help accelerate single-cell genomics workflows and analysis. It highlights how GPUs can speed up processing by around 20x compared to CPUs. The webinar demonstrates how Code Ocean compute capsules containing RAPIDS and ATACworks toolkits can provide a flexible computational environment for researchers to perform GPU-accelerated single-cell analysis on large datasets in a reproducible manner. This helps address challenges of growing dataset sizes and complexity for computational researchers.

Uploaded by

Namrata Lakhani
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 17

Accelerating Single-cell

Emily So
Genomics & Discovery with Solution Scientist

NVIDIA, AI and Code Ocean Code Ocean

1
Introduction

Emily So
Solutions Scientist, Code Ocean;
MSc. Candidate in Medical
Biophysics: Computational
Biology, University of Toronto

2
What will you learn at this webinar:

• Advantages & limitations of current AI/ML Usage


• Using GPU acceleration to scale and speed processing
and improve the accuracy
• Removing productivity limitations using Code Ocean
Compute Capsules
• Demonstration
‐ RAPIDS GPU Capsule
‐ ATACworks Capsule

3
GPU can accelerate Single Cell Analysis ~20X
CPU GPU

Source: NVIDIA Developer


- Accelerating Genomics
using RAPIDS

4
Audience Poll

5
Workflows in Single-Cell Genomics

- Explore genome-wide expression data at the single-cell level


- Increase in resolution is much better for analysis, using DNA/RNA
- Characterize cell populations
- Combine with other data for further applications (i.e. therapeutics, genetic disorders)
- Use of AI/ML has become gold standard for unsupervised clustering

6
Growing Single-Cell Datasets require AI/ML for Analysis
Success!
With AI/ML

Removal of
Dimensionality
Artifacts
Reduction

Batch Effect Single-Cell Data


Integration Analysis Machine Learning Accurate
Clustering
Results
Unsupervised
Clustering
Interactive,
Easy
Computational Environment Downstream
Analysis
7
Current Challenges

Trends Challenges Solutions

• ML algorithms in • Datasets continue to • Faster, GPU-accelerated


single-cell data analysis grow (> 1M cells) end to end workflows

• Able to produce clear, • Machine Resources • Flexible ML/Dev/Ops


accurate results on past
datasets • ML frameworks don’t • Researcher experience &
add speed operationalizing the team

• Increasing Complexity
for the Computational
Researcher

8
Today: How Computational Researchers can do this:

Trends Challenges Solutions

• ML algorithms in • Datasets continue to • Faster, GPU-accelerated


single-cell data analysis grow (> 1M cells) end to end workflows

• Able to produce clear, • Machine Resources • Flexible ML/Dev/Ops


accurate results on past
datasets • ML frameworks don’t • Researcher experience &
add speed operationalizing the
team
• Increasing Complexity
for the Computational
Researcher

9
Creating a Researcher Experience

Code Ocean Compute Capsule


Computational environment to package
NVIDIA CUDA, RAPIDS + AI/ML for
accurate, accelerated single-cell analysis

NVIDIA Clara Parabricks Toolkit


GPU acceleration of single-cell genomics
using RAPIDS and NVIDIA CUDA

AI/ML
Used in sequencing analysis as new
ways to produce efficient results for
growing single-cell datasets
AI/ML - NVIDIA accelerated for Genomic Analysis

10
Code Ocean Compute Capsule

Environment
- Self-serve UX for flexible compute resources,
code requirements

Code
- Multiple ways to execute models

- Easy to organize, share and reproduce

Data
- Easy to work with large datasets, uploading
from different sources

11
Live Demonstrations

RAPIDS GPU Compute Capsule

• GPU-acceleration on large datasets for single-cell analysis


• Datasets of 60k to 1.3M cells

ATACworks Compute Capsule

• Deep learning toolkit (with CUDA + RAPIDS) for coverage track


denoising and peak calling from low-coverage or low-quality
ATAC-Seq data.

12
What did we see
RAPIDS Researcher Experience Benefit:
• GPU acceleration Ready-to-use (Set-up)
ATACWorks - Ready for AI/ML
- Flexible machine resources
• GPU + RAPIDS
Productivity (Edit/Run/Share)
Code Ocean Compute Capsule
- Capture of results
• Computational Workstation - Share gene expression analysis
Quality (Management)
- UX for reproducibility

13
Questions?

14
Resources
• NVIDIA Clara Parabricks Toolkit
• developer.nvidia.com/clara-parabricks

• NVIDIA RAPIDS
• github.com/clara-parabricks/rapids-single-cell-examples

• NVIDIA ATACWorks
• github.com/clara-parabricks/AtacWorks

• Code Ocean Resources


• codeocean.com/resources
AI/ML - NVIDIA accelerated for Genomic Analysis

15
Next Steps

• Webinar recording to be provided

• Free access to GPU acceleration Example Capsules to be


provided

• Got questions? Contact: emily.so@codeocean.com

16
Thanks for attending!

Contact Code Ocean | Request a Demo 17

You might also like