A Review of Extremophile Protein Classification Using Machine Learning

Proteins can be denatured under extreme conditions due to breaking non-covalent bonds between protein side chains, causing conformational changes and loss of function. However, some proteins can withstand extreme conditions, known as extremophile proteins. The stability of this extremophile protein has attracted the interest of researchers to be used in industrial applications. Over the last decade, both traditional methods involve experimental and systematic comparison and computational methods such as machine learning are developed to classify extremophilic proteins and their counterparts and identify their essential properties as the basis for carrying out protein design and engineering. This review compares various studies to classify some extremophile proteins in terms of datasets, feature extractions, and classification algorithms used in each study. It also includes some challenges in this research area, such as data limitations, handcrafted features, limited sequence method utilized for classification, and model interpretability. The studies presented in this review are based on our unique review and by no means complete. Nevertheless, hopefully, the review will provide insight for new research in developing the more effective, interpretable and improved performance of predictions.