Related Experiment Video
Updated: Jan 18, 2026
08:04
Identification and Classification of Position-specific GABAA Receptor Subunit Missense Variants for Their Role In Hippocampal Pyramidal Neurons
Published on: June 6, 2025
1.4K
Predicting DDI-induced pregnancy and neonatal ADRs using sparse PCA and stacking ensemble approach
Anushka Chaurasia1, Deepak Kumar1, Yogita2
1Computer Science and Engineering, 385889 National Institute of Technology Meghalaya , Shillong, India.
Journal of Integrative Bioinformatics
|June 9, 2025
Summary
This study introduces a novel computational approach to predict drug-drug interaction (DDI)-induced adverse drug reactions (ADRs) in pregnancy and neonates. The method effectively handles sparse data and class imbalance, improving prediction accuracy for these critical populations.
Area of Science:
- Pharmacogenomics and Computational Toxicology
- Drug Safety and Pharmacology
- Maternal and Neonatal Health
Background:
- Predicting drug-drug interaction (DDI)-induced adverse drug reactions (ADRs) is complex due to limited data, sparsity, high dimensionality, and class imbalance.
- Existing computational methods for DDI-ADR prediction are primarily focused on the general population, leaving a gap in understanding risks for pregnant women and neonates.
Purpose of the Study:
- To develop and evaluate a novel sparse ensemble-based computational approach for predicting DDI-induced ADRs specifically in pregnancy and neonatal populations.
- To address challenges of high-dimensional, sparse data, and class imbalance in predicting these specific ADRs.
Main Methods:
- Utilized SMILES strings as molecular features for drug representation.
- Employed Sparse Principal Component Analysis (SPCA) to handle high-dimensional and sparse data, demonstrating improved performance over standard PCA.
- Applied the Multilabel Synthetic Minority Oversampling Technique (MLSMOTE) to mitigate class imbalance issues.
- Developed a stacking ensemble model integrating these techniques for final ADR prediction.
Main Results:
- SPCA showed a 2.67%-5.45% improvement in handling sparse data compared to Principal Component Analysis (PCA).
- The proposed stacking ensemble model significantly outperformed six state-of-the-art predictors.
- Achieved superior micro and macro scores for True Positive Rate (TPR), F1 Score, False Positive Rate (FPR), Precision, Hamming Loss, and ROC-AUC Score, with improvements ranging from 1.16% to 14.94%.
Conclusions:
- The proposed sparse ensemble approach effectively addresses data challenges in predicting DDI-induced ADRs for pregnancy and neonatal populations.
- This computational method offers a promising tool for enhancing drug safety assessments in vulnerable populations.
- The study highlights the potential of advanced machine learning techniques in specialized pharmacovigilance.
