Abstract <p>Drug discovery for any disease faces several challenges that impact the efficiency, cost, and success rate of developing new drugs. In silico methods have emerged as key tools for accelerating rational drug design and development, which helps save time and resources for predicting biological activities. Absorption, Distribution, Metabolism, and Excretion (ADME) - toxicity properties help in modelling the quality and quantity of drugs, as inaccurate data can diversify the chemical space. Therefore, selecting relevant molecular descriptors enhances the model’s performance. Severe Acute Respiratory Syndrome (SARS) is a contagious respiratory disease caused by the SARS Coronavirus (SARS-CoV). The proposed work aims to predict SARS bioactivities using a novel approach known as Convex Logistic Principal Component Analysis (CLPCA) based Quantitative Structure-Activity Relationship (QSAR) model. The three target proteins, namely SARS with ChEMBL_3927, ChEMBL_5118, and ChEMBL_4523582, are collected from the ChEMBL database. Five molecular fingerprints from the PaDEL descriptor have been selected to validate the performance of the proposed CLPCA-based QSAR model. R-squared and root mean square error (RMSE) are used to evaluate predictive performance. The proposed hybrid feature combination framework, which integrates CLPCA with the Support Vector Machine (SVR), achieved the highest R-squared value of 0.95 and a very low RMSE of 0.04. The in silico experiments with target proteins show that the proposed QSAR model performs better than the state-of-the-art methods. Thus, the proposed model increases the throughput for identifying novel inhibitors for treating SARS, achieving high predictive performance across multiple datasets.</p>