Closed darcyabjones closed 2 years ago
Hi Darcy, thanks a lot for raising this!
Turns out there was an issue in the conflict resolving function when processing Sec/SPII and Tat/SPII lipoproteins. I added logic to handle those as a separate case, using the predicted modified cysteine after the cleavage site to impute it when it's missing.
The online version is patched, I'll close the issue once the updated downloads go live.
Cool, thanks!
Hi there!
Thanks for your great work. I was testing out the update and came along an issue while running on the signalp5 benchmark set during the marginal conflict resolution step.
This sequence:
from https://services.healthtech.dtu.dk/services/SignalP-6.0/public_data/benchmark_set_sp5.fasta appears to be the issue.
Running version 6.0b in "fast" mode with this sequence in both other and eukaryote organisms causes the following error.
This doesn't appear to be an issue with the previous version available for download. Both have identical main dependency versions:
python 3.6.13 numpy 1.19.5 pytorch 1.9.1 tqdm 4.62.3
Thanks in advance, Darcy