Amyotrophic Lateral Sclerosis (ALS) is a progressive, fatal neurodegenerative disease which destroys motor neurons, causing muscular weakness, paralysis, and respiratory failure. In many cases of early-onset ALS, mutated Fused in Sarcoma (FUS) proteins are mislocalized from the nucleus to the cytoplasm, where it colocalizes with G3BP1, an RNA-binding protein, and forms aggregates known as stress granules (SG). Currently, models that predict localization states of G3BP1 and FUS1 within individual cells do not exist, with regular protein-protein interaction (PPI) networks yielding only averaged information across multiple cells. Our study aims to build dynamic PPI networks that are able to predict single-cell protein localization changes within ALS. As ground truth data, we will utilize pooled optical screening data whereby CRISPR knockouts were performed on cells and visualized for fluorescent FUS, G3BP, and DAPI as phenotypic perturbation markers caused by genetic mutation. Specific steps in our proposed methodology include taking pixel-based screening information to the latent space via a representation learner and training a perturbation predictor with the latent features matched with perturbation label data embedded via foundational gene/protein models. By building a latent feature predictor, we aim to create a model which can both validate current perturbations and extrapolate phenotypic predictions to new perturbations relevant to ALS pathogenesis. Overall, we hypothesize that such a model will inform us on which latent states are desirable for cellular survival, and which ones contribute to ALS disease progression.

I am participating in SuperUROP in order to combine my interest in molecular biology and big data with a deep-seated curiosity in the “how” of things. In this case, I hope to build a model that is not only able to predict phenotypes from changes in the genome, but uncover mechanistically why certain genetic mutations may be benign, while others lead to catastrophic disease. More broadly, I am excited to learn alongside my mentors and cohort, and see where interesting ideas lead me.