(answered) What is the expected run time for a Compound search ?

Started by Bletchley Park

Bletchley Park Lv 1

What should we expect performance -wise from that Compound lookup ?
Currently waiting for more than one and a half hours, this way this tool is useless.

rmoretti Staff Lv 1

It depends a bit on how many compounds are pending in the queue, but I would expect that you should typically get results within 5 minutes or so (probably less).

Several hours is not expected. If that happens, then there's different causes if it's just the one compound that's stalled, or if other submitted compounds are also stalled. If it's just the one, there's possibly something specific about that particular compound which the server search doesn't like. That should normally result in a "Failed" status, but there's potentially edge cases where it stalls at Running. If it's a large number of (very different) compounds, then there's likely something up with the server processing scripts, and the ferrets need to get nudged. (Or there's some connectivity issue with the foldit server itself.)

On Saturday we had a general outage, but as far as I'm aware, most new compounds should now be running, and there's only a handful of compounds which are specifically stalled.

LociOiling Lv 1

I had two that stalled at "running" over the weekend. Now they're both completed, with zero results. I resubmitted one, and it returned with zero results almost instantly. So I think that's a dead end.

I also resubmitted a previous one with 25 hits, and it quickly returned 25 hits again.

I'll try tinkering with the failed solutions to see what happens.

Bletchley Park Lv 1

How would we know that it has bad groups if it is not in the metrics ?
My solutions had all metrics at 1000.

I also find from a later succeeded lookup (with shorter comments of an improved version of the solution) that the orientations are completely different and it is hard to move things in a similar pose using bands. Is there a tutorial on how to rearrange the offered poses ? Are there any tools to achieve that or should we just submit 25 search results as-is ?

rmoretti Staff Lv 1

First off, we should have a fix for the stalled compounds, so you should get results from submissions (even if it's a no-results-found) in a reasonable timeframe. If you still have compounds listed as "Running" even after an hour, please let us know.

The library of compounds we're searching over with the current puzzle is the Enamine REAL set. This is an ultralarge make-on-demand library that Enamine thinks they can reasonably make. Only compounds on that list will get returned as results from the Compound Library. Which means that if the submitted compound isn't close to anything that Enamine thinks it reasonable to make, you won't get any results. Which compounds Enamine has on that list is orthogonal to the "bad groups" penalty, but there's potential overlap. Some of the bad groups are bad because they're difficult to make. As such, those sorts of groups won't show up in the library of compounds Enamine thinks they're able to make. So having a bad group or not isn't a determinate of getting results, but I wouldn't be surprised if compounds with bad groups are less likely to have a decent number of hits.

Regarding alignment, the alignment used for the overlay in the Compound Library panel should be based off the current ligand in the protein pocket. When you open the Load Library panel, it should re-compute that alignment based on whatever the current structure is. The problem you'll potentially run into is that the method used to align the compounds is currently less sophisticated than we'd like, so it may have a hard time with two structures which aren't chemically similar. You'll likely have to adjust things with Wiggle, Ligand Tweak, Pull and the move tool. (If you have suggestions about what sort of tools might make positioning in the pocket easier, we'd like to hear it.)

The other thing to keep in mind is that the Compound Library search results are based on the chemical identity of the compound (the atoms and bonding), and are independent of the 3D positioning of the ligand. Resubmitting the same ligand in a different conformation won't get you any different results. (It won't even affect the positioning of the substituted ligand.)

Bletchley Park Lv 1

Thank you for the elaborate reply.
"If you have suggestions about what sort of tools might make positioning in the pocket easier, we'd like to hear it.)"

If possible, please align the first and last atoms (structurally, '
If I design something like

A—I
B—I—B2
C—I

And I get D–E–F-G back, please align D with A and G with C in their original positions, the rest will hopefully wiggle itself out.