length of CDR1 region is not divisible by three

[williamdlees]

length of CDR1 region is not divisible by three

[marcelm]

Before running IgBLAST against the query sequences, IgDiscover runs IgBLAST on the V gene sequences alone in order to find where the FR1/2/3, CDR1/2 regions are located. That is, it delegates finding these regions to IgBLAST.

I believe the .aux file only contains information about the J genes, so that wouldn’t help here. (Also, IgDiscover forces IgBLAST to use an empty .aux file anyway.)

IgBLAST will honor the $IGDATA environment variable, which you can use to point to your own version of the data directory containing the internal_data/ directory with the necessary .ndm.imgt files. If you need a template, use the directory located at ...your-conda-installation-location/envs/igdiscover/share/igblast/ assuming your IgDiscover environment is named igdiscover. Be aware IGDATA needs to point to the directory containing the internal_data/ directory, not the internal_data/ directory itself.

The reason the length of the CDR1 regions are not divisible by three is probably that IgDiscover predicted a too short V sequence in a previous iteration, so another, probably easier way to solve the problem is to manually edit the iteration-xx/database/V.fasta file and add one (or two if needed) made-up nucleotides in the beginning of the problematic sequences.

IgDiscover should really not predict too short V sequences that it then trips over like this in the next iteration, but that is another issue.

[williamdlees]

Thanks, this is very helpful. In this case the short sequences were mine, not IgDiscover's.

[marcelm]

Ah great!