Identification of novel genes in Non-alcoholic fatty liver disease with Machine learning

Introduction: Non-alcoholic fatty liver disease (NAFLD) is defined as a range of chronic liverdiseases resulting from the accumulation of excess triglycerides in the liver. It is anticipated thatthe frequency of NAFLD will increase from ۸۳ million in ۲۰۱۵ to ۱۰۰ million by ۲۰۳۰. Diagnosisin the early stages reduces the risk of liver damage and increases the survival rate. Biomarkeridentification by computational approaches as reliable, and non-invasive methods is required toidentify specific diagnostic biomarkers in the early stage of fatty liver. Therefore, our aim wasto discover important genes related to Non-alcoholic fatty liver by machine learning algorithms.Method: The GSE۱۲۶۸۴۸ dataset included ۳۳۲۹۷ array-based expression profiling of ۷۳ samplesdownloaded from https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE۱۲۶۸۴۸. Dysregulatedexpression genes (DEGs) have been identified by machine learning methods (Penalizeregression models) based on Relief Weight feature selection after filtering and normalization. Theadjusted p < ۰.۰۵ and -۱.۵ <|Log۲FC (fold change) | < ۱.۵ were identified for subsequent analysisas significant genes. R۴.۱ and EVIEWS ۱۱ were used for analysis.Results: Elastic Net (ENET) was the robust predictor (Lambda at minimal error: ۱۱.۸۷, R۲=۰.۹۹۹and alpha = ۰.۵, l۱ Norm = ۱.۳۱). The area under the curve was approximately ۰.۹۹ with a confidenceinterval (۰.۹۵,۱). Four novel genes, including RABGAP۱, SLC۷A۸, SPAG۹, and KAT۶Awere found to have a differential expression between fatty liver and healthy individuals.Conclusion: The four key genes identified in our study. It is recommended that other prognosis,diagnosis and predictive biomarkers regarded to the fatty liver be discovered in further studies.

کلیدواژه ها:

Biomarker ، Gene ، Machine learning ، Fatty liver