Each bar denotes the log fold-change of the metabolite across the phenotype with a significant FDR-corrected test (q<0.5). Hover over the bars to see the p values, q values and test statistics. The downloadable CSV includes all tests (even non-significant ones). All fluxes are log-transformed and standardized so statistics are roughly comparable.
Download association test results as CSV...
The plot on the left shows a particular metabolite production flux by phenotype for significant
interactions. Those show individual metabolite associations with the phenotype. Hover over
individual points to see the metabolite annotations.
The plot on the right shows the overall cross-validation performance of
a regularized linear regression in explaining the phenotype. This uses a LASSO
regression with the regularization parameter learned from a 2-fold cross-validation.
The predictions themselves are all based on leave-one-out cross-validation. It can be
used as an estimate of the global assocaition of the phenotype with the flux state.
training accuracy = 1.0
cross-validation accuracy = 0.8 ± 0.0