Integration of metabolomics features using the network-based algorithm PIUMet

the identified m/z values that were significantly changed in all genotype comparisons, were uploaded to PIUMet (http://fraenkel-nsf.csbi.mit.edu/piumet2/). We additional included 60 features, which were unidentified but significant in all genotypes (Supplementary Table ⁴). We used the following parameters: number of trees 10, edge reliability 2, negative prize degree 0.0005, and number of repeats 50.

The Prize-Collecting Steiner Forest algorithm identifies metabolites and represents them as nodes, the higher the assignment score the bigger the node. The algorithm links these features based on high-confidence protein–protein and protein–metabolites interactions using two databases, HMDBv4.0 and Recon3D. Further details such as node frequency and node edge are reported in Supplementary Table ⁵. The output was processed using the R (Rstudio v.3.3.2) package “gplot” (v3.0.1) in order to visualize the cluster of metabolites and to highlight the connection between the predicted proteins and metabolites.