JellyByte
We build on foundational ML models able to predict the effects of perturbations on complex biological systems
Introduction
The integration of biology with machine learning (ML) stands at the forefront of scientific innovation, marking an era of significant breakthroughs and applications. This intersection, however, is currently experiencing a growth phase reminiscent of the early development stages of large language models (LLMs), characterized by its rapid evolution and the potential to revolutionize how we understand biological systems through data.
Problem
Drug discovery is an expensive and time consuming process in which most drugs never end up making it to the clinic. However, most drugs fail either because they do not induce the necessary effect (efficacy) or because they induce unwanted effects (toxicity). Given this fact, you would expect early-stage discovery to be guided by these two factors, but until now it has not been possible to take into account the long-range impacts of perturbations to a biological system as a whole that either cause reduced efficacy or increased toxicity.
Opportunity
We propose to change the "algorithm" that has defined drug discovery pipelines since the 1980s: target selection, hit discovery, lead optimization, ADMET, in vivo, and finally clinical trials. We are using AI models to incorporate long-range perturbation effects to guide the initial stages of this algorithm, thereby significantly improving the odds of drugs working in complex biological systems (such as humans!).