Metabolite-Group Selection On Temu Ireng (Curcuma Aeruginosa) Contains Related To Toxicity Activity By Using Group Lasso Regression

Metabolites are expressed in mass-to-charge ratio (m/z) on mass spectrometry experiments. They can be identified more than once. Some of m/z representing same metabolites can be considered as a group of metabolites. Evaluation of metabolite effects can be considered based on the groups. Group least absolute shrinkage and selection operator (group lasso) regression can be used to evaluate these groups. It shrinks some coefficients of regression exactly to be zero by adding intermediate penalty on ordinary least square (OLS) objective function. The purposes of this study were to estimate groups of metabolite contains of Curcuma aeruginosa (Temu ireng) affecting toxicity activity by using group lasso regression and to compare it to partial least square regression (PLSR). The data used were toxicity activity and metabolite contain, obtained from LC-MS, of temu ireng from three areas in Java. The groups of metabolites which affected toxicity activity, of group lasso regression by using dedicated software of R with gglasso package, were groups of m/z 238.150, 250.165, 262.128, 264.144, 312.275, and 456.183. The estimates of metabolites that affected of group lasso regression and PLSR had similarities. Based on the goodness of fit, group lasso regression was better than PLSR to estimate the affecting groups.