基于LASSO的非酒精性脂肪肝病的预测研究OA

Objective To develop a LASSO regression-based prediction model for non-alcoholic fatty liver disease(NAFLD)and identify potential disease-related genes to effectively distinguish patients from healthy individuals.Methods The GSE135251 dataset was obtained from the database(including 206 samples from NAFLD patients at different stages of fibrosis and 10 samples from healthy individuals),with all samples being transcriptome data obtained through high-throughput RNA sequencing.The Apriori algorithm was applied to analyze associations among gene loci and eliminate redundant or highly correlated features,thereby reducing data dimensionality.The LASSO regression algorithm was employed to build a NAFLD prediction model and identify potential disease-associated genes.Finally,based on the selected potential disease genes,the samples were grouped using the k-means unsupervised clustering method,and the consistency between the clustering results and the grouping of healthy individuals and NAFLD patients in the dataset was analyzed.Results The NAFLD prediction model identified PCBP2,CEBPD,GC,DNAJC12,and PTN as potential disease-associated genes.Based on these five genes,the samples were effectively clustered into healthy controls and NAFLD patients,yielding a silhouette coefficient of approximately 0.63 and a Davies-Bouldin index of about 0.5,indicating good cohesion and separation of the clusters.Conclusion The integration of association-based dimensionality reduction and machine learning enables effective identification of NAFLD-related genes and accurate sample classification,thus providing valuable insights for elucidating disease mechanisms and supporting potential clinical applications.