Trajectory numbers

[TimCraigCGPS]

In a previous thread it was mentioned that 200-300 trajectories are a good amount to run for "easy" targets, whereas "harder" targets may need 2000-3000. I am wondering what you normally consider a trajectory for this purpose.

Is a "trajectory" a line on trajectory_stats.csv or is it an entry in failure_csv.csv or perhaps the sum of lines in trajectory_stats and failure_csv?

The idea behind wanting to know this would be to be able to say that we have it a good shot and probably need to change hotspots or input models.

[martinpacesa]

A line in trajectory_stats.csv. Failure csv sort of tells you what the problem could be, such as many early trajectory plddt failures probably indicate bad starting model

[TimCraigCGPS]

I'm seeing this in my failure csv:

The input is an experimentally determined crystal structure. Also another thing I noticed- it seems like the colors on the box below the pLDDT structure doesn't match - What metric is being graphed in the box below pLDDT? Sometimes in earlier stages the boxes are shades of red/blue.

[martinpacesa]

oooh, yeah peptides are a bit harder, there you might need to sample even more, depending on the target. Perhaps try to even use the MPNN option to redesign the interface.

[TimCraigCGPS]

e the MPNN option to redesign the interface.

Thanks. Is this the parameter you're thinking of:

mpnn_fix_interface -> whether to fix the interface designed in the starting trajectory

[martinpacesa]

Yes!