ChAlPred

ChAlPred predicts the allergenic potential of chemical compounds using molecular descriptors and machine learning to classify chemicals as allergens or non-allergens.

Key Features:

• Dataset: 403 known allergenic chemical compounds and 1074 non-allergenic compounds were used for training, testing, and validation.
• Molecular descriptors: Use of 2D descriptors, 3D descriptors, and fingerprint (FP) descriptors including PubChemFP129, GraphFP1014, and KRFP890, with PubChemFP129 and GraphFP1014 more prevalent in allergenic compounds and KRFP890 more common in non-allergenic compounds.
• Machine learning models: Implementation of XGBoost (XGB), Random Forest (RF), and a hybrid model for classification tasks.
• Model performance: XGB achieved AUC 0.89 using 2D descriptors; RF achieved AUCs of 0.85 (3D), 0.92 (FP), and 0.93 (combined descriptors) on the validation dataset, and a hybrid model achieved AUC 0.92.
• Identified chemical signals: Predicted potential allergenicity for several FDA-approved drugs including Cefuroxime, Spironolactone, and Tioconazole.

Scientific Applications:

• Drug safety assessment: Prioritizes compounds for further evaluation in pharmacovigilance and drug development to identify potential chemical allergens.
• Environmental monitoring: Supports detection of potential allergenic chemicals in environmental samples and exposure assessments.
• Allergenicity research: Facilitates investigation of molecular features and fingerprints associated with chemical allergenicity.

Methodology:

Models were trained, tested, and validated on a dataset of 403 allergenic and 1074 non-allergenic chemicals using 2D, 3D, and FP descriptors (PubChemFP129, GraphFP1014, KRFP890) and classification algorithms including XGBoost, Random Forest, and a hybrid model with performance evaluated by AUC on a validation dataset.