Procedure for concatenating multiple trajectories (.npy format) from Adaptive sampling simulation

[SoumyaCYZ]

Hi @stefdoerr sir. I hope you are doing well. For TICA analysis of chignolin, I'm using the dihedral angle and pairwise distance as features. But I'm not getting the expected TICA. I've randomly concatenated many trajectories from adaptive sampling schemes. Maybe for this reason, I'm not getting the expected TICA plot. It'll be nice if you tell me the protocol for how I should concatenate the trajectories from the adaptive sampling. I have downloaded the trajectories from https://ftp.imp.fu-berlin.de/pub/cmb-data/

[stefdoerr]

Sorry for responding this late! I somehow skipped this issue. Using TICA with adaptive sampling is tricky and it depends on what software you used to perform adaptive sampling. In HTMD you can see the code which does it but it is specific to the naming scheme used in adaptive https://github.com/Acellera/htmd/blob/7ac9c8b4044d713cc44ea70d14a95d9ddca9a13f/htmd/adaptive/adaptive.py#L444
But since adaptive sampling can generate trajectory trees which is something TICA cannot really handle, you would have many trajectories with duplicated starts which could influence the statistics. The simplest thing would be to not join the trajectories, if they are long enough and use TICA on them without joining. We only ever used the function above for visualizing long trajectories from Adaptive, but never for actual MSM analysis