AMALPHI

AMALPHI predicts the potential of small molecules to induce drug-induced phospholipidosis (PLD) using machine learning models trained on curated chemical datasets.

Key Features:

• Machine Learning-Based Prediction: Applies machine learning classifiers to evaluate the phospholipidosis-inducing potential of small molecules.
• Balanced Random Forest Algorithm: Utilizes a balanced random forest model achieving predictive performance with an area under the curve (AUC) of up to 0.90.
• Curated Training Dataset: Trains models using a curated dataset of 545 small molecules derived from ChEMBL v30 with annotated chemical and biological properties.

Scientific Applications:

• Drug Safety Assessment: Predicts the likelihood that candidate compounds induce phospholipidosis during drug development.
• Antiviral Drug Repurposing Analysis: Helps distinguish true antiviral activity from phospholipidosis-associated effects in compounds evaluated against SARS-CoV-2.
• Medicinal Chemistry Optimization: Supports design of small molecules with reduced risk of inducing phospholipidosis.

Methodology:

AMALPHI trains balanced random forest machine learning models using a curated dataset of 545 small molecules from ChEMBL v30 to classify compounds according to their potential to induce phospholipidosis.