MEDICASCY

MEDICASCY predicts side effects, indications, efficacy, and protein modes of action of small-molecule drugs from chemical structure to support early-stage drug discovery and safety assessment.

Key Features:

• Input Requirement: Accepts a small molecule's SMILES string as the sole input representing chemical structure.
• Predictive Capabilities: Performs multi-label prediction of side effects, indications, efficacy, and protein modes of action from chemical structure alone.
• Algorithm: Implements a multi-label boosted random forest machine learning method.
• Performance Metrics: Retrospective benchmarking reports approximately 78% precision and recall for predicting at least one severe side effect and 72% precision for efficacy predictions.
• Experimental Validation: Efficacy predictions were experimentally validated on novel molecules, showing close to 80% precision in inhibiting the growth of ovarian, breast, and prostate cancer cell lines.

Scientific Applications:

• Early-stage drug discovery: Supports prioritization of candidate small molecules by predicting efficacy and indications from chemical structure.
• Safety profiling: Predicts potential side effects to assess the safety profile of novel compounds.
• Therapeutic target identification: Predicts protein modes of action to aid identification of potential therapeutic targets.

Methodology:

Uses a multi-label boosted random forest machine learning method that relies solely on chemical structure input provided as SMILES.