Classification of Biodegradable Chemicals Based on Their Structures and Physicochemical Properties Using Computational Simulations

The development and validation of reliable models for biodegradation are essential for us to assess the environmental impact of chemicals and to prioritize the development of sustainable chemicals. The biodegradability of a chemical depends on its chemical structure and physicochemical properties, such as molecular weight, solubility, and hydrophobicity. In the context of biodegradation, Quantitative structure-activity relationship (QSAR) models are used for this research to predict the biodegradability of chemicals based on their chemical structure and physicochemical properties. This prediction aids in the design of more environmentally friendly compounds.

In this paper, experimental values of 1055 chemicals were collected for the prediction. Forty-one molecular descriptors and one experimental class were used as attribute information. Explanatory Data Analysis (EDA) and data visualization were performed to identify the type and structure of the data and get the descriptive statistics. The studied biodegradation model involves selecting a training set of compounds with known biodegradability and identifying relevant molecular descriptors, such as electronic, topological, and steric properties. These descriptors with machine learning are then used to develop a mathematical model that can predict the biodegradability of new compounds.

It showed that the accuracy and reliability of QSAR models depend on the quality and completeness of the training data and the relevance of the molecular descriptors used in the model.