Fig. 2

Illustrative workflow to exploit a machine learning model trained on a large dataset (here, SMRT) to annotate metabolites. Steps 1-2: a RTs database is created using a predictive model trained on the SMRT dataset. Step 3: to use this database, a researcher provides the experimental RTs of a few molecules whose identity is known. The molecule identities are then used to retrieve the corresponding predicted RTs from the database to create pairs of experimental-predicted RTs. Step 4: a projection function mapping predicted RTs to experimental RTs is learned from these pairs. Step 5: the researcher then provides experimental m/z (not shown in the figure) of the molecules he/she is trying to identify. Molecules are filtered using the m/z ratio and the predicted RTs of those molecules are then projected to experimental RTs to create a “projected database”. Step 6: the researcher finally uses the experimental RTs to query the projected database. Step 7: the results retrieved from it would enable scoring candidates with similar m/z but different RTs