FP-MAP

FP-MAP predicts molecular bioactivities using fingerprint-based machine learning models to support drug discovery and lead compound identification.

Key Features:

• Extensive Model Library: FP-MAP contains approximately 4,000 classification and regression models evaluated on diverse bioactivity datasets.
• Diverse Disease Coverage: Models address diseases including neglected tropical diseases caused by viral, bacterial, and parasitic pathogens as well as Alzheimer's disease.
• Molecular Fingerprint Integration: FP-MAP employs 12 distinct molecular fingerprints to encode chemical structures for model training and prediction.
• Model Evaluation: Models were systematically evaluated for predictive performance on held-out test sets.
• Predictive Performance: Top-performing models report test set AUC values ranging from 0.62 to 0.99.

Scientific Applications:

• Hit identification and lead discovery: Predicting bioactivities to identify potential hit compounds and lead candidates across multiple targets.
• Screening and lead optimization: Supporting virtual screening and subsequent lead optimization by providing activity predictions across datasets.

Methodology:

Models use molecular fingerprints as inputs to machine learning classification and regression algorithms and were evaluated by test set AUC.