Abstract:
Understanding how molecular modifications influence biological activity is a cornerstone of drug discovery and development. This project leverages Positional Analog Scanning, a powerful technique in medicinal chemistry, to investigate how single-atom additions and functional group modifications impact the potency of lead compounds. By integrating advanced machine learning models such as Chemprop, this study uncovers complex structure-activity relationships (SAR) and statistical analyses to uncover patterns and predict biological activity changes. Morgan fingerprinting and external molecular descriptors are employed for feature engineering, while data preprocessing techniques address challenges such as skewed distributions and class imbalance. This work not only advances our understanding of SAR but also provides a scalable, data-driven framework to accelerate lead optimization, offering transformative potential for the drug development pipeline.
