|Anonymous | Login
|2024-03-05 07:37 CST
|My View | View Issues
|View Issue Details
|[All Projects] Crash
|Fixed in Version
|0000317: Growing loops can result in non-sensical loop sets and foldtrees.
|If you start things right, and have the right random seed, protocols::loops::Loops::grow_all_loops( ) can result in non-sensical loop sets. Specifically, it can result in loops that grow to be ontop of one another, e.g. a loopset which has both a loop from 27-33 and a loop from 27-35. There's a little bit of detection of this case in grow_loop(nres,loop,magL,magR), but it doesn't actually change anything or prevent it from happening.
The real kicker is when the extended, overlapping loop set gets passed to something like protocols::loops::fold_tree_from_loops(), where having overlapping loops and loops with coincident start sites makes an utterly repugnant foldtree.
It might be good to have some sort of normalization/merging code for the loop growing code (or really any loop set, which may be overlapping), which would normalize and merge loops such that they meet the standard conception of a loopset.
|An example which runs afoul with this using the protocols::comparative_modeling::LoopRelaxMover is posted at https://www.rosettacommons.org/node/3526 [^]
|No tags attached.
|Command Line Used
|Fixed in SVN Version
|Copyright © 2000 - 2012 MantisBT Group