Towards self-learning agents in era of high-throughput omics - - PowerPoint PPT Presentation

▶
towards self learning agents in era of high throughput
SMART_READER_LITE
LIVE PREVIEW

Towards self-learning agents in era of high-throughput omics - - PowerPoint PPT Presentation

Sep 17, 2023 383 likes •553 views

Towards self-learning agents in era of high-throughput omics Presenter: Ameen Eetemadi Principal Investigator: Prof. Ilias Tagkopoulos 1 I use Blue Waters to: 1. Design artificial neural 2. Determine optimal networks for gene strategies to

slide-1
SLIDE 1

Towards self-learning agents in era of high-throughput omics

Presenter: Ameen Eetemadi Principal Investigator: Prof. Ilias Tagkopoulos

1

slide-2
SLIDE 2

I use Blue Waters to:

2

  • 1. Design artificial neural

networks for gene expression prediction

  • Thermodynamic simulations
  • Deep Learning
  • Extensive evaluations
  • 2. Determine optimal

strategies to identify next set of experiments

  • Synthetic data generation
  • RNA-Seq data processing
  • Gaussian Processes
  • Extensive evaluations
slide-3
SLIDE 3

We have entered a new era …

3

High-throughput OMICS High-performance Computing Artificial Intelligence Robotic Equipment

Knowledge Discovery

m/z intensity
slide-4
SLIDE 4

Goal: Efficient Knowledge Discovery

4

  • E. coli Bacteria

Maximize knowledge about With Minimum Cost

Applications

Medicine Food Safety Basic Science

figure from: https://commons.wikimedia.org

slide-5
SLIDE 5

The Cycle of Knowledge Discovery

5

Learn Experiment

Data Lab

  • E. coli

Intelligent Agent

slide-6
SLIDE 6

Le Learn rn dynamic program of a cell

6

Le Learn rn dynamics of gene expression in a cell

Genetic Neural Network Key Features:

  • Captures Regulatory

Relationships

  • Models Transcription

Factor Dynamics

Published at: Bioinformatics Journal, 2018

Master Regulator

slide-7
SLIDE 7

Learn dynamics of gene expression in E. coli

7

Chemotaxis Transcription Regulatory Network

slide-8
SLIDE 8

Genetic Neural Network (GNN) is 40% more accurate (for chemotaxis genes)

8

Dataset size Mean Absolute Error Gene Expression (GE) Prediction Error

MLP RNN GNN BiRNN LinGNN Lasso

  • 0.10

0.15 10 40 70 100

Synthetic Data Used

slide-9
SLIDE 9

Genetic Neural Network (GNN) is 40% more accurate (for networks with 10-1000 genes)

9

GNN

Network size Mean Absolute Error Gene Expression (GE) Prediction Error

LinGNN MLP RNN BiRNN Lasso

  • 0.6

0.7 0.8 0.9 10 200 400 600 800 1000

GNN-rnd LinGNN-rnd

Real Data Used

slide-10
SLIDE 10

The Cycle of Knowledge Discovery

10

Learn Experiment

Data Lab

  • E. coli

Intelligent Agent

slide-11
SLIDE 11

Optimal Experimental Design for Gene Expression Prediction

11

slide-12
SLIDE 12

Accelerated Knowledge Discovery

12

slide-13
SLIDE 13

We have entered a new era …

13

High-throughput OMICS High-performance Computing Artificial Intelligence Robotic Equipment

Knowledge Discovery

m/z intensity

Genetic Neural Networks Optimal Experimental Design RNA-Seq Blue Waters Microarrays

slide-14
SLIDE 14

Blue Water Experience

Our Workload:

  • Extremely parallel
  • Independent small jobs

Advantages:

  • Extremely reliable
  • High availability
  • Comprehensive documentation

BW Customer Support:

  • Fast response
  • High quality

Thank You!

14

slide-15
SLIDE 15

Acknowledgements

  • Ilias Tagkopoulos, PhD (Principal Investigator)
  • Xiaokang Wang, PhD Candidate
  • Navneet Rai, PhD
  • Beatriz Merchel Piovesan Pereira, PhD Candidate

Funding:

15