Jeremy Wohlwend comments

Results 46 comments of


                                            Jeremy Wohlwend

Feature Request: Add --save_msa flag to export MSA files

Were you running with msa server when you ended up with an enmpty folder or custom MSA? I have verified that the folder is non-empty. Please share a way for...

Feature Request: Add --save_msa flag to export MSA files

@bio-rat we do not support automated templating. The goal of our templating logic is to allow people to use specific templates they want to use. Automated templating is something we...

affinity calculation for antibody-antigen models

We do not have a timeline to share at this time. I've seen some people have some success using the iPTM with antibodies, but it's not reliable.

Boltz-2 affinity prediction fails when using a CCD ligand

Should be fixed in the latest release and on the main branch. Please test and let me know! Make sure to delete your current output folder.

feat: add AF-sample2 style random MSA masking (--mask_msa)

Thanks for this contribution! Is the confidence model able to pick out that a better structure was found or does this only work under oracle setting?

feat: add AF-sample2 style random MSA masking (--mask_msa)

Cool, I’m good to merge with some documentation in prediction.md that also highlights this point! Would you be open to writing that?

Are pocket restraints with affinity prediction working?

Yep sorry we are fixing that. I ll update here when it’s done!

pLDDT reported in _ma_qa_metric_local loop as fraction rather than percent

@MattBrauer Are you sure about this? I thought the whole point of putting it in the b-factor was for mol* and other visualizers.

2.1.0 TypeError: Attention.forward() got an unexpected keyword argument 'biases'

My bad, this is fixed in 2.1.1. Btw your YAML has a problem, you can only pass a single binder to the affinity.

Out of memory error -- Nvidia A6000 48GB ~2300 aa residues

You can just do `CUDA_VISIBLE_DEVICES=1 boltz predict ...`