How does OmicsNet annotate MS peaks from untargeted metabolomics?

LC-MS peaks annotation is derived from the NetID algorithm. The whole process includes 5 steps:

  1. MS Peak Import: MS peaks from users’ uploading is recognized and cleaned. All MS signals meeting the ppm criteria and RT difference less 0.5% of the whole RT range will be merged as a single peak. All database and analyzing elements are loaded in this step;
  2. MS Data Pre-processing: All MS peaks are potentially matched towards HMDB database and annotated as multiple chemical identities basically; Besides, the intuitive relationships among the annotated peaks are found to establish a minimal network;
  3. Network Propagation: The network generated by the previous step will be propagated in this step. This propagation is trying to find out more associations among the nodes (peaks) as well as more potential chemical identities for each nodes. Twice bio-transformation together with three time abiotic propagation are performed;
  4. Global Network Optimization: The propagated network from last step is organized as a huge but sparse constraint matrix; The matrix is maximized with integer linear programming powered by lpsymphony package. As a result, only one chemical identity (node) and more confident bio-/abiotic relationship (edge) are kept;
  5. Annotation and Network Construction: The annotation results are being summarized in this step. Simultaneously, the network (for integration with other omics data) are filtered based on the p value cut-off of users’ input.