You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Describe the bug
Hello everyone, currently i am trying to index large peptide fasta files (~50 GB) for peptide searches. This fasta contains 85748938 entries of short peptides (all of them are unique). I am using the SABuild function and call it as follows:
I was wondering if this error could be fixed quickly, since i would like to use MSGF+ for identification, even for these large fastas i am using here. Maybe it is only a simple manner of using long instead of int, because of an possible overflow happening here. But i cannot judge if other places need to be adjusted.
This issue is caused by a limitation of the current implementation of MS-GF+ in Java, and fixing it is not a simple nor quick change. The issue is due to overflows on array sizes, and fixing it would involve changing arrays in many places to use an array type that supports indexing with long instead of int.
We do have other tools that we use for splitting fasta files into small enough sizes, then searching the data files with each fasta file, and then merging all results for a single data file back into one mzid file.
Thanks for the quick answer! and the clarification! Yes, splitting fasta files are always an option... However, i would look forward to execute a search via a single large fasta file.
If this is not a priority or not planned, then you can close this issue.
Activity
FarmGeek4Life commentedon Jul 1, 2022
Luxxii commentedon Jul 1, 2022
Thanks for the quick answer! and the clarification! Yes, splitting fasta files are always an option... However, i would look forward to execute a search via a single large fasta file.
If this is not a priority or not planned, then you can close this issue.