Deep Generative Learning of Single Cell Gene Expression to Predict Dose-Dependant Chemical Pertubations

Our webinar speaker is Sudin Battacharya, PhD, Associate Professor in the Departments of Biomedical Engineering and Pharmacology & Toxicology at Michigan State University. Dr. Battacharya’s research sits at the crossroads of computation and biology, where he uses quantitative tools to study how signaling and transcriptional networks control cell fate and how environmental pollutants can disrupt these processes. In this talk, he explores how artificial intelligence can be applied to single-cell gene expression data to predict dose-dependent chemical effects.

What you will learn:

How artificial intelligence can help predict how different cell types respond to chemicals at varying doses.

How a new tool, scVIDR, makes these predictions more accurate than previous methods.

How this approach can uncover the biological pathways driving these dose-dependent responses. Read the full abstract below.

Dr. Sudin Bhattacharya (he/him/his) is an Associate Professor in the departments of Biomedical Engineering and Pharmacology & Toxicology at Michigan State University. He leads a lab that conducts research at the interface of computation and biology, using quantitative tools to study the signaling and transcriptional networks that regulate cell fate and its perturbation by environmental pollutants. Dr. Bhattacharya joined MSU as an assistant professor in November 2015. Originally from Kolkata, India, Dr. Bhattacharya completed his undergraduate degree at Jadavpur University in India, his master’s degree at the University of Kentucky, and his Ph.D. at the University of Michigan, all in mechanical engineering. He did his postdoctoral training in computational biology at The Hamner Institutes in Research Triangle Park, North Carolina. His work has been funded by the National Institutes of Health and the US EPA.

Single cell RNA-sequencing allows us to study cell-type specificity and heterogeneity of biological responses to chemical perturbations. However, experimentally testing all relevant combinations of cell types, chemicals, and doses is a near-impossible task. A deep learning formalism called variational autoencoders (VAEs) has recently been shown to be effective in computationally predicting single-cell gene expression perturbations. We have developed single cell Variational Inference of Dose-Response (scVIDR), a VAE-based tool to predict the trajectory of cellular dose-response, which achieves better dose-response predictions than existing models. We show that scVIDR can predict dose-dependent gene expression changes across cell types, and interpret the latent space of the autoencoder model in terms of biological pathways.

ScitoVation helps clients assess chemical compound safety using innovative science, next-generation technology, and professional expertise. ScitoVation is known for partnership, flexibility, and proven success in its work to develop safer and more effective pharmaceuticals, food ingredients, agricultural chemicals, commodity chemicals and consumer products. A spin-off of the former The CIIT and The Hammer Institutes for Chemical & Drug Safety Sciences, ScitoVation is an industry leader of New Approach Methods (NAMS) for chemical/drug discovery & development in the rapidly evolving global regulatory landscape.