Predicting biochemical and physiological effects of natural products from molecular structures using machine learning

Area of Science:

• Natural Product Chemistry
• Cheminformatics
• Bioinformatics

Background:

• Advances in genomics and analytical techniques have accelerated the discovery of novel natural products.
• The growing volume of natural product data is crucial for developing machine learning (ML) models.
• ML models can analyze molecular structures to predict biological effects, aiding natural product research.

Purpose of the Study:

• To review recent studies on ML models for inferring molecular biological effects.
• To highlight the importance of molecular featurization in ML model development.
• To discuss technical challenges in applying ML to natural products.

Main Methods:

• Review of literature on ML models for natural product analysis (2016-2021).
• Focus on molecular featurization techniques for representing molecular structures computationally.
• Analysis of challenges in ML application for natural product research.

Main Results:

• Numerous ML models have been developed to predict various molecular properties.
• Molecular featurization is a critical step, influencing model performance.
• Significant progress has been made in understanding biological effects through ML.

Conclusions:

• ML offers powerful tools for predicting natural product biological effects.
• Effective molecular featurization is key to successful ML model development.
• Addressing technical challenges will further enhance ML applications in natural product discovery.