Development of a computational framework for the inference of protein variants in rare diseases amenable to pharmacological chaperones

Rare diseases represent a significant burden for public health systems and pose significant challenges to drug research and development. A promising strategy to address rare diseases is the repurposing of drugs already on the market. Conformational rare diseases, in particular, are treated with pharmacological chaperones, many of which are repurposed drugs. Despite the therapeutic potential of these compounds and the abundance of available literature, there is a lack of dedicated resources and databases enabling systematic data retrieval for research on rare conformational diseases. Moreover, although the literature is abundant, the molecular mechanisms underlying the activity of these compounds remain poorly understood, and, to our knowledge, no predictors of their efficacy on misfolding-associated variants are currently available. The main objective of my project is to implement Machine Learning (ML) models to infer protein variants associated with conformational RDs whose activity can be recovered via pharmacochaperones. Given the lack of resources to retrieve data on pharmacological chaperones used in the treatment of rare conformational diseases, I developed a platform aimed at researchers and medical doctors working in the field, in order to provide tools and data that might help gaining a better understanding of their usability and molecular mechanisms. During the first part of my PhD thesis work, I conducted an extensive bibliographic research to recover data on rare conformational diseases, the associated protein and protein variants, and the pharmacological chaperones tested for their treatment. Afterwards, I characterized protein variants associated with rare misfolding diseases with a computational pipeline. The data collected were then used to populate a relational database that was made public via the CIRCLE (Chaperones In Rare Conformational disEases) web platform, of which I have written both the frontend and the backend. Finally, from my database I have extracted a Machine Learning dataset that was central for the following work. In the second part of my PhD work I developed and implemented a pipeline to apply ML classification models and build a predictor that is being integrated in the CIRCLE platform. This tool will allow users to input a small molecule and predict its ability to recover the activity of variants associated with rare conformational diseases.