STRUCTURAL BIOLOGY MEETS PHARMACOGENOMICS: INSIGHTS INTO DRUG RESISTANCE

Abstract

Many cancers remain resistant to drugs and hence we require a good combination of approaches that are structural, molecular and computational.  As this research demonstrates, structural biology, pharmacogenomics, and machine learning are actually compatible, and all of them can collaborate to determine how drug resistance operates.  According to the high resolution molecular docking and dynamics analysis, some of these point mutations decrease binding affinities (DelG bind) and reduce the stability of protein-ligand complexes which can be evidenced by squared RMSD values and squared RMSF values.  IC50 was significantly larger in vitro tests on resistant and non-resistant cancer cell lines and major resistance genes (MDR1, BCRP, and CYP450) were over-expressed. qRT-PCR and western blot confirmed these results.  With pharmacogenomic data and the use of GDSC and CCLE, we identified additional key mutational hotspots, related to phenotype resistance, these include KRAS, TP53, and transporter A BC genes.  Structural and genomic variable-based ML models (Random Forest, XGBoost) were capable of predicting the resistance profile with an accuracy of 92. SHAP demonstrated that the greatest effect was on structural integrity and expression patterns.  The result of canonical correlation analysis found the statistically significant association between the changes in the molecular shape and genetic markers.  Each of the findings demonstrates that drug resistance is occasioned by various things which include changes in shape and alteration in transcription.  This integrative pipeline of methods provides us a means to predict and comprehend the mechanism of resistance, and is a foundation to the exact oncology strategies that attempt to remedy failure of therapy.