Master's Thesis

Sparse and Interpretable Models for Drug Mechanism-of-Action Discovery

Ana Rita do Nascimento Baião — 2022

Key information

Authors:

Ana Rita do Nascimento Baião (Ana Rita do Nascimento Baião)

Supervisors:

Emanuel José Vieira Gonçalves (Emanuel José Vieira Gonçalves); Susana de Almeida Mendes Vinga Martins (Susana de Almeida Mendes Vinga Martins)

Published in

November 10, 2022

Abstract

Precision medicine is increasingly used to personalize cancer treatments for each patient, providing tailored, effective, and less toxic therapies. Understanding drug mechanism-of-action is fundamental to improve drug development success rates. Recent work integrated anti-cancer drug sensitivity measurements with genome-wide CRISPR loss-of-function screens across cell lines to systematically investigate cellular drug mechanism-of-action. In this thesis work, linear regression sparse regularization models were developed to expand beyond the associations of a drug with a single gene, and integrate simultaneously thousands of genes. These novel methodologies integrate the penalties in Elastic Net models using tailored feature regularization metrics based on protein-protein interaction networks centrality measures and multi-omics integration. Notably, sparse regularization models selected at most only 2% of the total number of genes considered in the CRISPR-Cas9 screens for each drug. Thus, biologically interpretable solutions were obtained, with only a small subset of genes being important for each drug response prediction. Several associations between drugs and their canonical targets were recapitulated, and by leveraging PPI networks, biological and functional networks of proteins underpinning drug response were characterized. This revealed important information about the drug mechanism-of-action, such as drug potency, specificity, and potential off-target activities. Collectively, this thesis illustrates how linear regression sparse regularization models can be used to integrate cell line drug sensitivity with CRISPR loss-of-function screens to inform drug mechanism-ofaction to possibly guide the development of novel therapies.

Publication details

Authors in the community:

Supervisors of this institution:

Fields of Science and Technology (FOS)

industrial-biotechnology - Industrial Biotechnology

Publication language (ISO code)

eng - English

Rights type:

Embargo lifted

Date available:

September 8, 2023

Institution name

Instituto Superior Técnico