STTR Phase I: A Machine Learning Toolbox to Identify Therapeutics for Rare Genetic Disorders from Phenotypic Screens on Micropattern-Based Organoids

The broader impact/commercial potential of this Small Business Technology Transfer (STTR) Phase I project is to develop a widely applicable, innovative pipeline for drug discovery against rare genetic disorders. While each orphan disease affects a small fraction of individuals, thousands exist, affecting 400 million people world-wide. The challenge to bring therapies to the clinics for diseases that currently have no cure, which is coupled with an expanding market. Orphan drug sales are expected to reach $262 billion in 2024. Over the last few years, organoids made from human embryonic stem cells have raised the possibility of understanding the mechanisms leading to a genetic disorder by reconstituting, in vitro, a similar arrangement of the specific cell populations that are implicated in a certain disease. By comparing organoids made from healthy cells to those carrying the associated genetic mutation, drug testing may be performed in vitro to discover new therapeutics. However, this potential has not been harnessed due to technical limitations. In this proposal, as a necessary step towards this goal, the plan is to develop the analytical tools based on machine learning algorithms that will allow the analysis of drug therapeutic potential and toxicity when applied to organoid cultures. The intellectual merit of this STTR Phase I project is to develop and validate an unbiased analytical scheme to quantify the results of high-throughput phenotypic screens performed on micropattern organoids. Organoids have the potential to revolutionize the pre-clinical part of the drug discovery pipeline as they provide unbiased endpoints, they are disease relevant, and sensitive to toxicity. However, the use of multi-cellular organization in a high-throughput platform requires analysis of large amounts of information-rich microscopy images. The tools developed in this proposal will allow turning the biological complexity of tissue scale phenotypes into information predictive of the clinical potential of a compound. In the first aim, the goal is to demonstrate that the proposed method allows quantifying compound toxicity. The plan is to calibrate the toolbox against known methods quantifying cell viability, applying known toxic agents and withdrawn drugs. This will constitute a stringent test for the efficiency of the proposed method at quantifying deleterious effects of a compound. Second, the goal is to optimize the algorithm to measure therapeutic potential, using positive controls on micropattern-based organoids modeling Huntington's Disease (HD). The objective is to demonstrate the ability to analyze results of a screening campaign aimed at phenotypic reversal on the HD organoids, and detect hit molecules from a small library of FDA-approved drugs. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.