Executive Summary : | The project intends to develop a novel method combining cheminformatics with deep learning to unravel the complex mechanisms of drug-drug interactions (DDIs). DDIs are undesirable drug reactions that occur via simultaneous consumption of two or more drugs. The last decade has seen a surge in polypharmacy for diseases like cancer, infections (SARS-CoV-2, Mycobacterium tuberculosis), and neurodegenerative (Alzheimer’s). The underlying concept is to target multiple factors causing the disease, leading to a more effective cure. Polypharmacy is frequent among elderly, who intake multiple drugs to deal with different health conditions. This is particularly grave in India. Such strategies cause DDIs, which are difficult to analyse during clinical trials. These interactions may be lethal, e.g., interactions between warfarin and antimicrobials, amiodarone and statins, PDE-5 inhibitors and nitrates. That is why the computational understanding of DDIs is of utmost importance. This genre has immense scope of investigations. DDIs mainly occur during drug-metabolism, primarily caused by hepatic cytochrome P450 (CYP) isozymes in humans including CYP1, 2 and 3. DDIs also appear if two drugs share similar structure, common target or pathway. Enormous human CYP structural data is present in Protein Data Bank; their orderly dynamics for sufficient timescale and role of mutations and folding stability change in altering drug interactions is still unresolved. Here, we propose a systematic study on CYP dynamics, point mutations in CYP and drug interactions with four crucial classes of drugs: antibiotics, antiviral, anti-neurodegenerative and anticancer. Their choice is guided by the prevalence of polypharmacy in these conditions. To our best knowledge, no major study shows interactions with antibiotics using deep learning. The novelty also lies in quantifying CYP dynamic features in drug interactions. We intend to use the drug and natural product databases for DDIs analysis. We will adopt a multidimensional approach integrated with the state-of-the-art graph neural network deep learning that include: (a) insight of the CYP isozymes via molecular dynamics simulations and their structure-function relations, (b) differential features finding of CYP substrates, inhibitors and inducers, (c) CYP protein folding stability analysis upon mutations, (d) incorporating CYP structure simulation, binding site and genetic parameters with drug physicochemical, empirical, target, phenotypic and substructure features for DDIs prediction tool development. This method is not specific to the four selected drug categories but can be generalised to other diseases as well. Our approach will be further strengthened by active in-house collaborations with Data Science and Artificial Intelligence discipline. The project addresses the complexities of polypharmacy in a comprehensive yet general way. It will foster novel ideas and concepts in computational methods tailored to simulate DDIs. |