Identifying Potential Natural Antibiotics from Unani Formulas through Machine Learning Approaches.

Area of Science:

• Integrative Medicine
• Computational Biology
• Pharmacology

Background:

• Unani Tibb, a Greek-origin medical system, is prevalent in South and Central Asia.
• Unani herbal medicines, derived from plants, are used in primary healthcare.
• Growing antibiotic resistance and superbugs necessitate novel therapeutic strategies.

Purpose of the Study:

• To investigate Unani herbal ingredients for potential natural antibiotic properties.
• To address antibiotic resistance, multi-drug resistance, and superbug emergence.
• To identify molecular-level effects of Unani compounds using machine learning.

Main Methods:

• Employed 12 machine learning algorithms, including decision trees, kernels, neural networks, and probability-based methods.
• Utilized data preprocessing techniques: Synthetic Minority Over-sampling Technique (SMOTE), Feature Selection, and Principal Component Analysis (PCA).
• Optimized machine learning models through grid-search tuning of hyperparameters.

Main Results:

• The Multi-Layer Perceptron (MLP) model with SMOTE preprocessing achieved high accuracy, precision, and recall.
• Identified 20 crucial metabolites with predicted natural antibiotic activity.
• Validated predictions through literature review and structural similarity analysis with known antibiotics.

Conclusions:

• Unani herbal medicine constituents show promise as novel natural antibiotics.
• Machine learning effectively predicts antibiotic potential of natural compounds.
• This research offers a data-driven approach to discovering new antimicrobials against resistant pathogens.