Why are MSReact (crest) created structures so central for QCxMS2?

[tobigithub]

Why is MSREACT so central to QCxMS2. From reading the QCxMS2 paper it seems that MSREACT is used to create input structures.

• MSREACT [Paper] and
• MSREACT code [LINK].

Basically (there might be multiple naïve questions and statements), it works for generating starting fragments for EI and CID, and those fragments are used for input using transition state theory calculations to obtain energetically feasible structures, by exploring a reaction network of possible reactions.

But what happens if MSREACT does not create such feasible reaction products, or for example if the time domain is different, this could be femtoseconds (EI) to microseconds (CID). Also for example complex reactions like a Retro-Diels-Alder reaction may not be captured correctly by the initial MSREACT run. Is this all recovered during the reaction network exploration? Is it crucial that MSREACT covers all the initial fragments, or does the secondary step of exploring many reactions along the PES account for that.

Assuming that MSREACT gives us completely wrong results, would QCxMS2 still recover from that? Or if MSREACT is completely correct with fragments, will QCxMS2 compute the spectrum with high quality?

Thank you!
Tobias

[gorges97]

The fragment generation step is crucial for the entire QCxMS2 workflow. If we miss important fragments at this step, there is no way to recover them later, as reaction barriers and rate constants are only calculated for these fragments.
The purpose of using GFN2-xTB constrained optimizations in MSREACT is to remain on a reasonable potential energy surface (PES) and generate all relevant fragments while avoiding generating all possible fragments to keep the total number of computations feasible.

Your example of the Retro-Diels-Alder reaction is indeed challenging for MSREACT. I just tested it for cyclohexene, and the corresponding fragment at m/z = 54 is only generated when using extended settings of additional random atom shifts (I used "-msnshifts 100"). We are currently working on improving MSREACT for such complex reactions, particularly those involving ring structures and rearrangements.

[tobigithub]

Awesome thanks @gorges97!