Knowledge Base

When docking a large database, why does Glide sometimes fail with an error like the following? posesort(): error 256 executing glide_sort command; will try 1 more time Out of memory! Is there a workaround, or a way to recover the intermediate results?

When docking a large database, it is important to split the job into multiple subjobs, so that the individual subjobs aren't too large. Very large subjobs may dock all the structures successfully, but the final sorting step is done in memory, which effectively limits the number of poses per subjob to 50-100K, depending on the amount of available memory.

For a 1 million compound library, we suggest splitting the job into at least 20 subjobs, and even more if you are expanding the number of ligand states with LigPrep or saving multiple poses per ligand. In addition to avoiding memory limitations, this will allow individual subjobs to complete in a reasonable amount of time, which is better for stability (very long processes are more susceptible to a variety of problems), and efficiency in restarting (fewer repeated calculations if a subjob has to be rerun). You can run a distributed Glide job with many more subjobs than processors/licenses used; these are independent parameters chosen in the Start dialog (or via '-NJOBS njobs' and '-HOST host/queue:ncpus' from the command line). Using more subjobs than processors also can be helpful for load balancing.

To ask a question or get help, please submit a support ticket or email us at help@schrodinger.com.

Back To Top