Novel Predictive Machine Learning Approach for Identification of Microbial Niche and Microbial Communities from Omics Dataset of Kaveri River, Tamil-Nadu, India

The Micro-plastics are a consequence of plastic pollution. They are harmful to both humans and the environment. They cause some disorders such as oxidative stress, DNA damage and inflammation, among other health problems to living organisms. Hence, need to remove them from the environment. Degradation and remediation techniques currently available to decompose micro-plastics also pose some challenges to researchers. The structuring of microbial communities is governed by microbial niches and stochastic processes. Therefore, this study proposes a novel predictive machine-learning approach for identifying microbial communities using the physicochemical properties of microorganisms that inhabit MP. The Kaveri River, located in Tamilnadu, India was used as a case study. The decision tree-based neural network, Principal Component Analysis (PCA), rank-based feature selection, decision tree, and single hidden layer feed-forward network were applied. From the results obtained, the decision tree method achieved 74% of accuracy with a minimum number of physicochemical attributes of micro-organisms while the decision tree-based neural network method achieved the highest mean accuracy of 96.67%. The outcomes of the results suggest that the decision tree-based neural network has the potential to be used as a predictor system for novel micro-organisms. Hence, the new model approach will be useful for the remediation of MP in our environment and the ecosystem.