Anton producedoutput for metabolite identification MS3 level identification is generated either by MS2 query and Siruis There are several metrics to judge the level 3 identificaiton - MS2Query_model_prediction: score is witin 0-1. The bigger the better: model score > 0.7 often yields very good information; model score between 0.6 and 0.7 are sketchy but might yield information; model score < 0.5 is not trustworthy at all. - MS2Query_mz_diff: dPPM: The difference of mz between measured compound and .... The smaller the better, 0.0003 is considered as small enough that it's probably exactly that compound; 22.02 is huge and it's not exactly that compound, it's just similar to it. - Sirius conf: Siruis confidence score: The bigger the better. For example, 0.298 is a low sirius score, 0.397 is middle and 0.736 is high Using these metrics to select either MS2query or sirius identification - If MS2Query_model_prediction is high and MS2Query_mz_diff is very small ------->MS2Query level 2 annotation - If MS2Query_model_prediction is high and MS2Query_mz_diff is not so big ------->MS2Query level 3 annotation - If MS2Query_model_prediction is high and MS2Query_mz_diff is very big ------->Not identifiable - If MS2Query_model_prediction is low ------->Not identifiable - If Sirius conf is high -------> Sirius level 3 annotation - If Sirius conf is low ------->Not identifiable It is possible that even if the counpund should have MS2Query annotation, we cannot find the long name of the compound in the database, e.g, .... PubChem, archkey... Then it is not identifiable cf_direct_parent: Anton could you explain this part? How feature clusters is processed: Anton could you explain this part?