NPAtlas

NPAtlas maps and classifies microbial natural products by chemical structure to analyze structural diversity and predict bacterial or fungal origin using MAP4, TMAP, and machine learning.

Key Features:

• MinHashed Atom Pair fingerprint (MAP4): NPAtlas computes MAP4 fingerprints (MinHashed atom-pair fingerprint with a diameter of four bonds) to represent molecular structures across a wide range of sizes for structural diversity analysis.
• Tree Map dimensionality reduction (TMAP): NPAtlas applies TMAP to reduce dimensionality and visualize chemical space, organizing molecules by physico-chemical properties and compound families such as peptides, glycosides, polyphenols, and terpenoids.
• Machine learning model: NPAtlas trains a machine learning classifier to distinguish bacterial versus fungal origin of natural products based on structural features derived from MAP4 and TMAP representations.

Scientific Applications:

• Drug Discovery and Development: Classifying and mapping microbial natural products aids prioritization and identification of novel candidate molecules for therapeutic development.
• Structural Biology and Chemistry: Analysis of MAP4-derived representations and TMAP visualizations provides insights into structural diversity and physico-chemical properties of natural products.
• Biosynthetic Pathway Analysis: Distinguishing bacterial and fungal natural products supports interpretation of biosynthetic origins relevant to synthetic biology and metabolic engineering.

Methodology:

NPAtlas extracts MAP4 molecular fingerprints, applies TMAP for dimensionality reduction and chemical-space visualization, and trains a machine learning model on these data to predict bacterial versus fungal origin.