I am using Haddock to compare docking simulations of mutants of a protein with peptides.
I am using the same AIRs (NMR derived binding residues for each docking job). Although I have found run to run I get slightly variable results. In that haddock scores and cluster, sizes, etc differ.Even if I submit the same files with the same options (with random AIR exclusion turned on as a default). Just trying to understand what I am doing wrong or could change to remedy this?
I have been considering if the Random AIRs are actually hindering me in my particular use I know I can turn them off but I first wanted to understand if I am understanding correctly how they work. Each time I submit a job I specify a list of say 10 AIRs the random exclusion default means 50% are excluded so AIR 1,5,7,9,10 in 1 submission but in the next it could be 2,5,7,8, and 9. Or do the random exclusions vary within each job?
Are you using the web server? Jobs are distributed on a worldwide grid of computers. You will only get exactly the same results if you run on exactly the same hardware, which is not the case.
So small fluctuations are to be expected - but the overall picture should remain consistent.
In some cases the differences are more than slight. I am comparing binding orders as in a recent covid ace2 example. I am using the web server. I am thinking of trialling with AIR random exclusion off but wanted to check I understood correctly?
Are you using the web server? Jobs are distributed on a worldwide grid of computers. You will only get exactly the same results if you run on exactly the same hardware, which is not the case.
So small fluctuations are to be expected - but the overall picture should remain consistent.
I was using the scores as an indication of the order of binding is this not viable? Can you provide a link or reference outlining how to use the refinement interface? I am needing to dock to get a structure to submit. I think I need to review my strategy…
Thanks I think I now understand. I will dock each of my peptides with my WT protein to get a WT complex. Then introduce mutations into the WT complex PDBs. Then use the refinement interface to see if the mutations have any effect on the scores for each peptide complex variant. I will turn off the random AIR for each full dock though as I want the same residues to be used.