Cell Painting predicts impact of lung cancer variants

Juan C. Caicedo, John Arevalo, Federica Piccioni, Mark-Anthony Bray, Cathy L. Hartland, Xiaoyun Wu, Angela N. Brooks, Alice H. Berger, Jesse S. Boehm, Anne E. Carpenter, Shantanu Singh

What is this about?

Clinical sequencing uncovers many variants of unknown significance, especially in cancer, leaving doctors stuck. We show that cell morphology一specifically, images of cells stained for various organelles through the Cell Painting assay一contains sufficient information to determine the impact of many variants on many genes’ functions. We collected Cell Painting images of A549 cells for 325 overexpression constructs of variants found in lung adenocarcinoma (LUAD) across 50 reference genes in a high-throughput experiment. Then, we used image-based profiling to transform the Cell Painting images into cell morphology data that can be analyzed to identify phenotypic differences in an unbiased way. Next, we adapted the Variant Impact Phenotyping (VIP and VIP2) algorithm to make predictions using the cell morphology data; we call it cell morphology-based VIP (cmVIP) . Finally, we conducted various data analyses to validate and compare the Cell Painting predictions with transcriptional profiling (L1000).

We found that Cell Painting reveals insights into the phenotypic characteristics of cancer variants, which are generally consistent with transcriptional profiling. We created this website to allow researchers to explore the predictions and results of most of the analyses performed in our project. This website is designed to share all the results obtained in the project in a highly interactive way.

Resources

  • Here is a link to the preprint in bioRxiv with more details of this study.
  • The full Cell Painting image collection created for this project can be found here.
  • This GitHub repository contains all the code and instructions to reproduce the analysis presented here.

Instructions

The 50 reference genes are listed in the left panel of this website. Click on one gene to access the list of corresponding variants tested in this project.

When clicking on one variant, the cmVIP predictions will appear with links to single cell results as well as L1000 (eVIP) results.

The top menu presents a list of options to various global views of the dataset and comparisons with transcriptional profiling.