Toward Repurposing Metformin as a Precision Anti-Cancer Therapy Using Structural Systems Pharmacology

Abstract

Metformin, a drug prescribed to treat type-2 diabetes, exhibits anti-cancer effects in a portion of patients, but the direct molecular and genetic interactions leading to this pleiotropic effect have not yet been fully explored. To repurpose metformin as a precision anti-cancer therapy, we have developed a novel structural systems pharmacology approach to elucidate metformin’s molecular basis and genetic biomarkers of action. We integrated structural proteome-scale drug target identification with network biology analysis by combining structural genomic, functional genomic, and interactomic data. Through searching the human structural proteome, we identified twenty putative metformin binding targets and their interaction models. We experimentally verified the interactions between metformin and our top-ranked kinase targets. Notably, kinases, particularly SGK1 and EGFR were identified as key molecular targets of metformin. Subsequently, we linked these putative binding targets to genes that do not directly bind to metformin but whose expressions are altered by metformin through protein-protein interactions, and identified network biomarkers of phenotypic response of metformin. The molecular targets and the key nodes in genetic networks are largely consistent with the existing experimental evidence. Their interactions can be affected by the observed cancer mutations. This study will shed new light into repurposing metformin for safe, effective, personalized therapies.

VariFunNet, an integrated multiscale modeling framework to study the effects of rare non-coding variants in genome-wide association studies: Applied to Alzheimer's disease

“…. We quantify the impact of node inhibition on the network robustness using Natural Connectivity [38]. The Natural Connectivity corresponds to an “average” eigenvalue of a graph.…”

Comparing Drug Images and Repurposing Drugs with BioGPS and FLAPdock: The Thymidylate Synthase Case

“…More recently, approaches combining both chemical and target information have been reported showing that a drug action is often unspecific, and underlying the necessity of combining biology and chemistry to provide reliable molecular explanations for complex SEs . Ligand binding site comparison and protein–ligand docking have been also successfully integrated and applied for drug repurposing, side effect prediction and polypharmacology applications …”

“…[34] Ligand binding site comparisona nd protein-ligand docking have been also successfully integrated and applied for drugr epurposing,s ide effect prediction and polypharmacologya pplications. [35][36][37][38] Approachesd irectly comparing protein pockets have also been recently proposed, based on the assumption that similar binding sites can be targeted by similarl igands, [39][40][41][42][43] and that the structuraland chemical informationencoded into the binding pockets guide the recognitionb etween macromolecules and ligands. [44][45][46][47] SMAP is af ast method forl igand binding site comparison, using as hape description only based on Ca atoms.…”

Identifying protein subsets and features responsible for improved drug repurposing accuracies using the CANDO platform

“…Incorporating machine learning, which is continuing to prove its utility in 41 many aspects of biomedicine [18][19][20] including drug discovery and repurposing [21,22], 42 August 24, 2018 2/17 into the CANDO platform to increase benchmarking accuracies and therefore its 43 predictive power is of importance. Various algorithms can be incorporated (for example, 44 neural networks, support vector machines, and decision trees), but the well documented 45 issues described by the curse of dimensionality [23,24] will plague any choice in the 46 current (v1) implementation of CANDO, especially considering the extremely large 47 number of features (≈ 50,000 proteins) within each compound-proteome interaction 48 signature vector. Given the relatively few training samples (an average of ≈ 9 drugs per 49 indication), a machine learning approach to train how drugs interact with proteomes is 50 a much easier task with a vastly reduced set of proteins.…”