The Role of Machine Learning and Cell-Painting in Drug Development​

The Role of Machine Learning and Cell-Painting in Drug Development 09.09.2020 | Fachbereich Chemie | Luis Vollmers | Master Thesis | University of Cambridge 1

What is Cell Painting? What is Cell Painting? 09.09.2020 | Fachbereich Chemie | What is ML? Luis Vollmers How to use ML and CP? | Master Thesis | Why is that relevant? University of Cambridge Questions? 2

What is Cell Painting? What is Cell Painting? What is ML? How to use ML and CP? Why is that relevant? Questions? vs. 09.09.2020 | Fachbereich Chemie | Luis Vollmers | Master Thesis | University of Cambridge 3

Treatment, Staining and Plating What is Cell Painting? 09.09.2020 | Fachbereich Chemie | What is ML? Luis Vollmers How to use ML and CP? | Master Thesis | Why is that relevant? University of Cambridge Questions? 4

High-Throughput-Imaging What is Cell Painting? 09.09.2020 | Fachbereich Chemie | What is ML? Luis Vollmers How to use ML and CP? | Master Thesis | Why is that relevant? University of Cambridge Questions? 5

Automated Analysis of the Cell Images What is Cell Painting? What is ML? How to use ML and CP? Why is that relevant? Questions? • CellProfiler • Recognizes Cells • Recognizes Organelles • Generates Numerical Data • Downstream Data Analysis (ML) 09.09.2020 | Fachbereich Chemie | Luis Vollmers | Master Thesis | University of Cambridge 6

Data Set Used in My Master Thesis What is Cell Painting? What is ML? How to use ML and CP? Why is that relevant? Questions? • ~1800 features extracted by CellProfiler • Features are AreaOfNucleus, ERGranularity, CellRoundness, etc… • ~30000 distinct compounds • The total data set comprises ~54000000 entries • No human could ever comprehend so much information • Machine Learning (ML) Methods can use high-dimensional Information 09.09.2020 | Fachbereich Chemie | Luis Vollmers | Master Thesis | University of Cambridge 7

What is Machine Learning? What is Cell Painting? 09.09.2020 | Fachbereich Chemie | What is ML? Luis Vollmers How to use ML and CP? | Master Thesis | Why is that relevant? University of Cambridge Questions? 8

Traditional Programming: a.k.a. Software 1.0 What is Cell Painting? 09.09.2020 | Fachbereich Chemie | What is ML? Luis Vollmers How to use ML and CP? | Master Thesis | Why is that relevant? University of Cambridge Questions? 9

Traditional Programming: a.k.a. Software 1.0 What is Cell Painting? 09.09.2020 | Fachbereich Chemie | What is ML? Luis Vollmers How to use ML and CP? | Master Thesis | Why is that relevant? University of Cambridge Questions? 10

What Software 1.0 is really bad at What is Cell Painting? 09.09.2020 | Fachbereich Chemie | What is ML? Luis Vollmers How to use ML and CP? | Master Thesis | Why is that relevant? University of Cambridge Questions? 11

Machine Learning: a.k.a. Software 2.0 What is Cell Painting? 09.09.2020 | Fachbereich Chemie | What is ML? Luis Vollmers How to use ML and CP? | Master Thesis | Why is that relevant? University of Cambridge Questions? 12

Different Machine Learning Models What is Cell Painting? What is ML? How to use ML and CP? Random Forest Classifier Why is that relevant? Questions? Support Vector Machines Neural Networks 09.09.2020 | Fachbereich Chemie | Luis Vollmers | Master Thesis | University of Cambridge 13

The Simplest Deep Neural Network What is Cell Painting? 09.09.2020 | Fachbereich Chemie | What is ML? Luis Vollmers How to use ML and CP? | Master Thesis | Why is that relevant? University of Cambridge Questions? 14

The Simplest Deep Neural Network What is Cell Painting? 09.09.2020 | Fachbereich Chemie | What is ML? Luis Vollmers How to use ML and CP? | Master Thesis | Why is that relevant? University of Cambridge Questions? 15

Calculating the Hidden Layer: h What is Cell Painting? 09.09.2020 | Fachbereich Chemie | What is ML? Luis Vollmers How to use ML and CP? | Master Thesis | Why is that relevant? University of Cambridge Questions? 16

Calculating the Output: y What is Cell Painting? 09.09.2020 | Fachbereich Chemie | What is ML? Luis Vollmers How to use ML and CP? | Master Thesis | Why is that relevant? University of Cambridge Questions? 17

How to Combine CP and ML? What is Cell Painting? What is ML? How to use ML and CP? Why is that relevant? Questions? What would be the target? 09.09.2020 | Fachbereich Chemie | Luis Vollmers | Master Thesis | University of Cambridge 18

Toxicity Endpoints are Key What is Cell Painting? 09.09.2020 | Fachbereich Chemie | What is ML? Luis Vollmers How to use ML and CP? | Master Thesis | Why is that relevant? University of Cambridge Questions? 19

Open Source Toxicity Databases (DB) What is Cell Painting? What is ML? How to use ML and CP? Why is that relevant? Questions? • Examples for Open Source DB • DILIrank – Livertoxicity • DrugBank – Complex Encyclopedia with various Endpoints • EcoTox – Data on Ecological Toxicity • SIDER – Side Effects Data Base (Adverse Drug Reactions) • PubChem - Complex Encyclopedia with various Endpoints • Handling of Data bases is not trivial • Size, accessibility and complexity of the stored information • Data needs to be understood before it can be used in a model 09.09.2020 | Fachbereich Chemie | Luis Vollmers | Master Thesis | University of Cambridge 20

Why is Toxicity Prediction from CP beneficial? What is Cell Painting? 09.09.2020 | Fachbereich Chemie | What is ML? Luis Vollmers How to use ML and CP? | Master Thesis | Why is that relevant? University of Cambridge Questions? 21

Thank you for your attention! Feel free to ask questions. What is Cell Painting? 09.09.2020 | Fachbereich Chemie | What is ML? Luis Vollmers How to use ML and CP? | Master Thesis | Why is that relevant? University of Cambridge Questions? 22