1 Predictive metabolomic profiling of microbial communities using amplicon or metagenomic sequences
开发的软件名字叫做MelonnPan(Model-based Genomically Informed High-dimensional Predictor of Microbial Community Metabolic Profile)
在一个cohort上开发, 在另一个indepent dataset上做验证.
This revealed high concordance (一致) between predicted and observed community metabolic trends in >50% of metabolites whose identities were confirmed against laboratory standards, including prediction of metabolic shifts associated with bile acids, fatty acids, steroids, prenol lipids, and sphingolipids.
1.1 Methods
1.1.1 Training and validation cohort descriptions
One training data sets, two validation data sets.
1.1.2 Taxonomic and functional profiling
Stool samples.
1.1.3 Metabolite profiling
HHILIC and RPLC. 8869 features and 466 metabolites.
1.1.4 Elastic net regularization
Elastic net regularization technique. Use the R package glmnet.
Spearman correlation coefficient (r) between the true and pre- dicted metabolite compositions was used to evaluate the predictability of each compound. Following Cohen68, we term those metabolites for which r is >0.3 as “well predicted” and flag the rest as poorly predicted metabolites.
1.1.5 Significance testing with shuffled data
其实就是permutation test.
