InflamNat

InflamNat provides a curated database and machine learning–based predictive models to evaluate anti-inflammatory natural products and predict compound–target relationships for drug discovery applications.

Key Features:

• Extensive Database: Contains detailed information on 1351 natural products including physicochemical properties, cellular bioactivities, and molecular targets.
• Predictive Tools: Provides two machine learning–based predictors: one for anti-inflammatory activity of natural products and one for inferring compound–target relationships for compounds and targets lacking existing relationship data.
• Innovative Methodology: Implements a multi-tokenization transformer (MTT) as the sequential encoder in both predictive tools to generate high-quality sequential representations.
• Performance Metrics: Experimental validation reports model performance evaluated by Area Under the Curve (AUC).

Scientific Applications:

• Screening of natural products: Prioritizes natural product candidates for anti-inflammatory activity using predicted activity scores and database annotations.
• Target identification: Infers potential compound–target relationships to support identification of molecular targets for natural products.
• Drug discovery prioritization: Supports selection and prioritization of NP candidates in preclinical stages of anti-inflammatory drug development.
• Data-driven characterization: Enables analysis of physicochemical properties, cellular bioactivities, and molecular targets to inform experimental hypothesis generation.

Methodology:

Machine learning–based predictive models tailored for natural products employ a multi-tokenization transformer (MTT) as the sequential encoder to predict anti-inflammatory activity and infer compound–target relationships, with experimental evaluation reported using AUC.