Grant: CIHR Fellowship

Pharmaceutical companies, agrochemical companies, and government agencies detect chemical toxicity with animal testing, which is expensive, slow, ethically concerning, and often inaccurate for predicting human health, prompting scientists to develop non-animal testing methods. One idea is to grow human cells in a dish, expose them to chemicals, measure detailed biological data, and predict whether real-life concentrations of that chemical could be toxic in humans. However, it is difficult to translate impacts in isolated cells to impacts in complex organisms and while these new methods do work to some extent, each fails to capture many toxicity outcomes for many chemicals. My goal is to determine if this problem can be solved by growing the cells more like real organs, or by collecting more diverse data. One data type that seems promising is image-based profiling, where pictures taken with high-powered microscopes are analyzed to quantify the size, shape, and texture of components inside of thousands of individual cells. I will analyze imaging data from increasingly complex cell models exposed to compounds known to cause liver toxicity to determine which toxicity outcomes are detected in basic cell models versus co-cultures (where two or more types of cells are grown together; 60 compounds), organoids (where cells are grown in 3D instead of 2D; ~500 compounds), and organ-on-a-chip models (where microfluidics are used to simulate the inside of an organ; ~500 compounds). For a larger group of compounds (~1600), I will analyze image data from the basic cell models, along with the expression of ~3000 genes and the abundance of ~200 proteins, allowing me to determine how well different data types detect different types of toxicity outcomes. While these questions have long been of interest, the technology is only now becoming mature enough to investigate them at scale. The findings of this research could inform testing methods used around the world for decades to come.