AISM-REGBF: An Interpretable Regression-Based Framework for Predicting Small- Molecule Potency

Abstract Inflammation is a primary immune response categorized into acute and chronic types, with chronic inflammation linked to various diseases such as autoimmune disorders and cancer. Conventional therapies often face limitations due to side effects and therapeutic resistance, prompting the exploration of small-molecule compounds as safer alternatives. In this study we emphasize the integration of machine learning (ML), particularly through structure-activity relationship (SAR) analysis and quantitative structure-activity relationship (QSAR) modeling, to enhance drug discovery. A series of ML-based regression models, including LightGBM, were developed and systematically evaluated. Various feature selection method, were employed to effectively reduce the input dimensionality and improve model performance. The LightGBM model, based on combined Morgan and substructure count fingerprints, achieved strong predictive performance with an R² of 0.75 on the independent test dataset. External validation with FDA-approved compounds affirmed the model's applicability, highlighting key structural features identified through explainable artificial intelligence (XAI) techniques, specifically SHAP and LIME. This study concludes that the AISM-REGBF framework provides a reliable computational tool for predicting the anti-inflammatory potency (pIC₅₀) of small molecules.

Authors

Institutions

Publication Details

Journal
Research Square
Published
2026-11-16
DOI
https://doi.org/10.21203/rs.3.rs-11464167/v1
Primary Topic
Computational Drug Discovery Methods
Type
preprint
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
preprint

AISM-REGBF: An Interpretable Regression-Based Framework for Predicting Small- Molecule Potency

Thirumurthy Madhavan, Priya Dharshini Balaji, Subathra Selvam
3 links
Research Square
Computational Drug Discovery Methods
preprint

AISM-REGBF: An Interpretable Regression-Based Framework for Predicting Small- Molecule Potency

Thirumurthy Madhavan, Priya Dharshini Balaji, Subathra Selvam
preprint en

Abstract

Abstract Inflammation is a primary immune response categorized into acute and chronic types, with chronic inflammation linked to various diseases such as autoimmune disorders and cancer. Conventional therapies often face limitations due to side effects and therapeutic resistance, prompting the exploration of small-molecule compounds as safer alternatives. In this study we emphasize the integration of machine learning (ML), particularly through structure-activity relationship (SAR) analysis and quantitative structure-activity relationship (QSAR) modeling, to enhance drug discovery. A series of ML-based regression models, including LightGBM, were developed and systematically evaluated. Various feature selection method, were employed to effectively reduce the input dimensionality and improve model performance. The LightGBM model, based on combined Morgan and substructure count fingerprints, achieved strong predictive performance with an R² of 0.75 on the independent test dataset. External validation with FDA-approved compounds affirmed the model's applicability, highlighting key structural features identified through explainable artificial intelligence (XAI) techniques, specifically SHAP and LIME. This study concludes that the AISM-REGBF framework provides a reliable computational tool for predicting the anti-inflammatory potency (pIC₅₀) of small molecules.

Research Square
SRM Institute of Science and Technology (IN)
Good health and well-being
Computational Drug Discovery Methods
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon