ReCuRandom: A hybrid machine learning model for significant gene identification.

The method of evaluating the physical traits of living creatures by generating essential proteins is known as gene expression. Translation and transcription are the two processes in gene expression. Gene Expression can be captured from DNA or RNA using a variety of approaches. High-throughput gene expression (NGS) data is having multiple advantages over the traditional gene expression dataset. Still one of the major disadvantages of NGS data is it is complex, comprising a large number of genes with few samples; among these thousands of significant genes, disease-causing genes must be identified. This issue should be solved by lowering the data source's dimension to a manageable level. Machine learning (ML) has proven its potential in the research areas of disease identification, and genomics in recent years. Many Machine Learning-based Gene Selection techniques have been described in the literature, to improve dimensionality reduction precision. In this work, a unique clinical support system based on the Relief-Cuckoo Search (CS)-Random Forest (RF) hybrid multi-stage significant gene identification technique is proposed for significant gene selection from the next-generation sequencing (NGS) dataset. This work can be helpful for the accurate identification of sepsis and the relevant genes associated with it, which helps early detection of the disease. This hybrid feature selection technique could be beneficial in later stages of disease treatment, such as disease diagnosis and drug development. The accuracy of the proposed model is 95.23 % which can be helpful for further analysis of various infectious diseases and thus helps the diagnosis procedure more convenient. [ABSTRACT FROM AUTHOR]