Few issues about MSDIAL and MSFINDER May 17, 2020, 06:13:42 PM Hi, here are some issues and my own expectations for MSDIAL and MSFINDER. MS-DIAL (data independent LC-MS/MS) a. When dealing with Waters MSE data, in my experience, it is better to remove function 3 files (_FUNC003.DAT, _FUNC003.IDX, AND _FUNC003.STS, they are for the lockmass) from the *.raw file before converting into *.abf file for MSIDAL processing. Otherwise, the deconvoluted MS/MS spectra might weird. b. There is always an error "library information should be imported". But I have MSP library downloaded from RIKEN, and use it for the identification. Is it only available when the spot hit any library? c. I am wondering if MSDIAL allows manually identification of any feature since the requisite of uploading spectra to MoNA is that the feature has to be fully identified. If I wanna submit the spectrum of a compound, I know which spot is the compound, but how can I assign metadata to that spot and then submit it? And I think it will be useful to allow to export the identified spot as local database, e.g. *MSP file. d. Can anybody tell me what is correlation based deconvolution? I could not find useful information to fully understand what is it for. MSFINDER a.Is it possible to the selected rather than all local databases for formula scoring? I think it will be nice for research areas other than metabolomics. b. Is it possible to allow us to mark down the probable correct structure we think? It seems like the "reflect MS-FINDER result to MSP/MAT file" function. The difference is that it would be better to permit us to select the one we want instead of to choose always the top 1 candidate. And it would be very nice to allow exporting the "identified" compound (meta data) together with MS/MS spectrum as a MSP file. It will facilitate building library and then submit to open library since metadata are already there. Using predicted retention time to structure elucidation is really a good idea. However, is it possible to allow user to integrate their well train model (for example trained and tested by Retip R package) for the prediction instead of using the simple XLogP method?