PREDICTING BIOLOGICAL ACTIVITY ON TARGETED THERAPEUTICS WITH THE INTEGRATION OF MACHINE LEARNING AND VIRTUAL SCREENING FOR PDAC DRUG REPURPOSING

Abstract

Owing to pancreatic cancer, and its aggressiveness, low efficiency of therapy, and issues with early identification, pancreatic cancer ranks among the most common causes of cancer-related mortality in medical research. For the purpose of improving the diagnosis and treatment of pancreatic ductal adenocarcinoma (PDAC), it is thus essential to find new diseaseassociated genes and targets. In this study, we attempted to identify the differentially expressed genes (DEGs) of PDAC using machine learning algorithms, performed virtual screening to repurpose drugs, and made biological predictions based on different bioac-tivity studies. Initially, we pre-processed the microarray gene expression data of PDAC and normal, publicly available in the NCBI database using a standard pipeline. Then, we identified the DEGs using machine-learning algorithms such as extreme gradient boosting (XGBoost) and random forest (RF). Further, we performed the virtual screening on the identified DEGs to repurpose the drugs using the DEPMap database. Finally, we analyzed the potential inhibitory roles of identified drugs on genes using ADMET profiling, drug-likeness, and bioactivity. Our results revealed that ERCC3, ACY3, ATP2A3, MW-TW, and ZBTB7A were common in the top 25 genes ranked by XGBoost and RF. We found five drugs such as Phenelzine, Doramapi-mod, ICI-162846, Afatinib, and Axitinib for the ERCC3 gene with a high positive correlation value of above 0.35. Further, we found that Axitinib has the highest biological activity which is active for G-protein coupledreceptor, kinase, protease, and enzyme inhibitors and moderately active for ion channel modulators and nuclear receptor ligands. The findings pave the way for new directions in targeted treatment approaches for PDAC, with possible wider applicability across other cancer types. © 2024 IAE All rights reserved.