Automating High-Content Screening Image Analysis with Deep Learning - - PowerPoint PPT Presentation

Automating High-Content Screening Image Analysis with Deep Learning

GPU Technology Conference May 10, 2017 Oren Kraus, University of Toronto

1

High-Content Screening (HCS)

2

Arrayed chemical or genetic conditions Automated microscopy Cellular phenotypes

Applications

3

Image based drug profiling

(Gustafsdottir, 2013) (Chong, 2015)

Functional genomics

Why HCS?

4

Fitness Morphology Phenotype

Cellular fitness as colony size

Fish hook spindle

(mcm21∆)

Wild-type spindle

Growth defect

(Vizeacoumar, 2010) (Costanzo, 2016)

Previous approaches

5

(Breker, 2013)

Classifying images by eye Measuring specific parameters

(Zanella, 2007) (Singh, 2014)

Previous implementations of HCS relied on manual inspection of each image

• r extracting specific measurements for a particular study

Phenotypic profiling

6

Single cell segmentation Feature extraction & selection Phenotypic profiles Classification Clustering & visualization

(Kraus, 2016)

Phenotypic profiling is used to construct quantitative and unbiased representations of single cells for classification or clustering

Complexity of profiling pipelines

7

(Boutros, 2015)

• Imaging protocol needs to

be optimized together with analysis pipelines

• Segmentation needs to be

tuned for every screen

• Feature extraction &

selection needs to be

• ptimized for every task
• Machine learning algorithm

needs to be tuned for every task

• Analysis pipelines are not

transferable across screens

Deep learning ‘solves’ object recognition

8

Deep learning has significantly outperformed feature extraction based classifiers on the popular ILSVRC image recognition benchmark since 2012

(Kraus, 2016) (Russakovsky, 2015)

Deep learning basics

9

Updated during training Updated during training Prediction compared to label

Nuclear periphery (Kraus, 2017)

Deep learning features

10

(Kraus, 2016)

Quantifying proteome dynamics with HCS

11

Fluorescently tagged strain generation Protein sub-cellular compartment localizations Localization changes in response to RPD3 genetic deletion

(Chong, 2015)

Ensemble of 60 binary SVMs (EnsLoc)

12

(Chong, 2015)

DeepLoc: Classifying protein localization with deep learning

13

Convolution Non-Linear Activation

Convolutional Block Max Pooling Fully Connected Layers Output Input

(Kraus, 2017)

DeepLoc is a deep convolutional neural network consisting of 11 layers and is trained to classify images of single cells into 17 localization categories

DeepLoc classification performance

14

(Kraus, 2017)

DeepLoc significantly outperforms ensLoc at single cell protein localization and at annotating the localization of proteins

DeepLoc features

15

DeepLoc

(Kraus, 2017)

DeepLoc EnsLoc

The single cell profiles learned by DeepLoc are better at representing cell images compared to extracted features

How DeepLoc works

16

Input Samples Convolutional Filter Visualizations (Layer 8) * = Layer 8 Activations

(Kraus, 2017)

Features in DeepLoc are activated when a learned pattern matches the input image

17

Abundance Log2 Change

Untreated Factor Treated Fdo1 Mcd1 Kss1 Yor342c Prm1 Prm2 Fus1 Nvj1 Bfa1 Group 1 Group 2 Group 3

Discovering proteins involved in mating

(Kraus, 2017)

Using DeepLoc on other images

18

Eisosomes Endosome ER Golgi Bud Tip Bud Site Bud Neck Actin

Chong 2015 Labels New Labels

(Kraus, 2017)

Transfer learning with DeepLoc

19

Using only 5 labeled samples per class, DeepLoc can be fine-tuned to accurately assess protein localization for most categories

(Kraus, 2017)

Deep learning automates HCS analysis

Deep learning automates the analysis of microscopy images by combining the steps in previous pipelines into one transferable model

20

(Kraus, 2017)

Acknowledgments

Boone and Andrews Labs (Ben Grys) Frey Lab (Jimmy Ba)

21 Jimmy

22

Deep Learning High Content Screening Discovery & Diagnostics

www.phenomicai.com