کنفرانس مهندسی برق ایران

صفحه اصلی / سی و چهارمین کنفرانس بین المللی مهندسی برق

Comparative Analysis of Machine Learning and BBI-ConvSE for Biopharmaceutical Drug Interaction Prediction

نویسندگان :

Fatemeh Nasiri¹ Mahdi Nouroozi² Mohsen Hooshmand³

1- دانشگاه تحصیلات تکمیلی علوم پایه زنجان 2- دانشگاه تحصیلات تکمیلی علوم پایه زنجان 3- دانشگاه تحصیلات تکمیلی علوم پایه زنجان

کلمات کلیدی :

Drug-drug interaction،Biotech drug،Machine learning،Deep learning،BBI-ConvSE،Squeeze-excitation،Random Forest،Predictive modeling

چکیده :

Drug-drug interactions (DDIs) are a major concern in modern pharmacotherapy, as they can lead to reduced therapeutic efficacy or severe adverse effects. While numerous computational approaches have been developed for predicting interactions among small-molecule drugs, biotech drugs remain understudied despite their growing clinical importance. This work introduces a novel dataset and a comprehensive comparative framework for predicting interactions involving biotech drugs. We benchmark traditional machine learning (ML) methods against two deep learning architectures—a standard Convolutional Neural Network (CNN) and our proposed \textbf{BBI-ConvSE}, a hybrid model integrating squeeze-excitation attention for adaptive feature recalibration. Contrary to trends in other domains, our evaluation reveals that ensemble ML methods (Random Forest, XGBoost) consistently outperform deep learning approaches across multiple class-imbalance scenarios. This finding highlights that for the current scale of biotech drug data, feature-engineered ML models provide superior predictive accuracy. Our study provides practical guidance for model selection in this emerging domain and establishes a performance baseline for future research in biopharmaceutical safety.