A Comparative Study using Next Generation Sequencing Data and Machine Learning Approach for Crohn's Disease (CD) Identification

Abstract

Multi-Drug resistance organisms (MDRO) are a major threat to the world. Crohn's disease (CD) is a chronic inflammatory disorder caused by bacteria, which mostly affects the gastrointestinal tract and is complex and heterogeneous. Generally, antibiotics are prescribed to treat CD but, overuse or misuse of antibiotics may lead to AMR. The trait's genetic architecture is mainly unknown. One of the medical genetics' main goals is to reliably predict CD based on these genetic and environmental factors, which may further help in the research of Antimicrobial resistance (AMR). The traditional approach for identifying the important genes that are causing the disease is complex with less accuracy. This complexity forces the researchers to search for an alternative, mainly focused on developing a prediction model with more accuracy and cost-effectiveness. The advancement in computational intelligence especially machine learning (ML) can handle the process of identifying the significant genes of this trait in an efficient and more biological significant manner. In this work, we use the recent growth of ML methodologies to identify the significant genes associated with CD. The effect of quality control (QC), imputing, and coding methods on non-linear model results revealed that QC and imputation of missing genotypes can artificially boost scores. The linear approach is found with an accuracy average of 0.52 whereas the accuracy average obtained with the non-linear model is more than 0.60. In addition to this, advanced high-throughput sequencing techniques like next-generation sequencing (NGS) help to improve the quality of data for analysis and thus lead to a better result. The ML approach especially the non-linear methodology like the extra tree classifier (ET) performs extremely well to find the CD with less computational time and high accuracy.

Topics & Concepts