NPs (Natural products)

NPs classifies natural products by origin using MAP4 (MinHashed atom pair fingerprint) descriptors and a Support Vector Machine trained on the COCONUT database to support biosynthetic source identification and drug discovery.

Key Features:

• Data Source: Utilizes the COCONUT database, an extensive repository of natural products with known origins.
• Fingerprinting: Uses MAP4 (MinHashed atom pair fingerprint) to represent molecular structures for similarity analyses.
• Clustering and visualization: Clusters and visualizes natural products according to origin using MAP4-based similarity measures.
• Classification model: Implements a Support Vector Machine (SVM) trained on MAP4 fingerprints, reporting accuracies of 94% for plant-derived NPs and 89% for both fungal and bacterial NPs.

Scientific Applications:

• Drug Discovery: Identifies the biological origin of natural products to streamline the search for novel compounds with therapeutic potential.
• Biosynthetic Gene Research: Aids in pinpointing biosynthetic genes associated with NPs isolated from plants that may be produced by endophytic microorganisms.

Methodology:

Generates MAP4 (MinHashed atom pair fingerprint) representations for molecules, uses MAP4-based clustering and visualization by origin, and trains a Support Vector Machine (SVM) on MAP4 fingerprints with reported accuracies of 94% (plants) and 89% (fungi and bacteria).