Integrating Artificial Intelligence and Data Science for Breakthroughs in Drug Development and Genetic Biomarker Discovery

Md Habibur Rahman; Md Abubokor Siam; Ahmed Shan-A-Alahi; Kazi Bushra Siddiqa; Shuchona Malek Orthi; Md Kazi Tuhin; Emran Hossain; Mukther Uddin

doi:10.63332/joph.v5i8.3157

Authors

Md Habibur Rahman School of Business, International American University, Los Angeles, CA 90010, USA
Md Abubokor Siam College of Business, Westcliff University, Irvine, CA 92614, USA
Ahmed Shan-A-Alahi Department of Technology and Computer Science, University of The Potomac, Washington DC, USA
Kazi Bushra Siddiqa School of Business, International American University, Los Angeles, CA 90010, USA
Shuchona Malek Orthi College of Business, Westcliff University, Irvine, CA 92614, USA
Md Kazi Tuhin Katz School of Science and Health, Yeshiva University, 245 Lexington Avenue, New York, USA
Emran Hossain Department of Business Administration, Humphreys University, Stockton, California, USA
Mukther Uddin School of Business, International American University, Los Angeles, CA 90010, USA

DOI:

https://doi.org/10.63332/joph.v5i8.3157

Keywords:

Artificial Intelligence, Drug Development, Genetic Biomarkers, Biomarker, Data Science, Genomics of Drug Sensitivity in Cancer (GDSC).

Abstract

The evolving complexity of drug discovery and the demand for focused therapeutics have given further impetus to the implementation of AI and data science-based approaches. Such methods can analyze genomic and pharmacological data more rapidly, facilitating the precision design of drugs and genetic biomarkers of such drugs. This work is a machine learning framework that seeks to predict drug responses and find genetic biomarkers in the Genomics of Drug Sensitivity in Cancer (GDSC) dataset. Numerous data cleansing, normalization and one-hot encoding were also done to maintain credibility in the analysis. The robustness of a Random Forest classifier on the processing of high-dimensional biological data and excellent predictive performance was also demonstrated, with 97.7% accuracy, 98.4% precision, recall, and F1-score. The comparative studies provided better results than other models like SVM 95%, BiLSTM 80%, and GATv2 77.9%. The discriminative power of the model was proved using ROC and precision-recall curves. The framework of AI + data science in pharmacogenomics can be used to identify patterns of drug sensitivity efficiently, and, thus, promote personalized medicine and biomarker-based treatments. The method is a scalable, interpretable, and time-saving alternative to the traditional pipelines of drug discovery.