ADEMA: An Algorithm to Determine Expected Metabolite Level Alterations Using Mutual Information

摘要: Metabolomics is a relatively new “omics” platform, which analyzes discrete set of metabolites detected in bio-fluids or tissue samples organisms. It has been used diverse array studies to detect biomarkers and determine activity rates for pathways based on changes due disease drugs. Recent improvements analytical methodology large sample throughput allow creation datasets that reflect metabolic dynamics perturbation the network. However, current methods comprehensive analyses (metabolomics) are limited, unlike other approaches where complex techniques analyzing coexpression/coregulation multiple variables applied. This paper discusses shortcomings metabolomics data analysis techniques, proposes multivariate technique (ADEMA) mutual information identify expected metabolite level with respect specific condition. We show ADEMA better predicts De Novo Lipogenesis pathway Cystic Fibrosis (CF) than prediction significance individual changes. also applied ADEMA's classification scheme three different cohorts CF wildtype mice. was able predict whether an unknown mouse genotype 1.0, 0.84, 0.9 accuracy each respective dataset. results had up 31% higher as compared algorithms. In conclusion, advances state-of-the-art analysis, by providing accurate interpretable results.