Bug MHC class 2 - Clustering did not include all 1000 input models

[Danielle-314]

A TCR and pMHC were docked with command line: python3 /home/ddiepenbroek/swifttcr/scripts/swift_tcr.py -r /projects/0/prjs1135/dkfz/swifttcr/input/short_seq/ss_tcr118_m3_on_pmhc_m0/pMHC_shortseq_tcr118_mut1_dr_m0.pdb -l /projects/0/prjs1135/dkfz/swifttcr/input/short_seq/ss_tcr118_m3_on_pmhc_m0/TCR_shortseq_tcr118_mut1_dr_m3.pdb -o /projects/0/prjs1135/dkfz/swifttcr/output -op ss_tcr118_mut1_dr_m3_m0 -c 16 -t 9 -m 1000 -mhc2

The clustering file clustering.txt contained a total of 997 models by counting the total amount of center models and neighbours. I rerunned the clustering for 6A clustering_6A.txt with the following python script in swifttcr: swifttcr/scripts/clustering.py and four models were missing. The models missing for 6A are ['merged_679.pdb', 'merged_898.pdb', 'merged_908.pdb', 'merged_992.pdb'] and for 9A ['merged_679.pdb', 'merged_908.pdb', 'merged_992.pdb'] . Clustering is performed with the following file: irmsd.csv.

[JanAarts]

I have checked your irmsd file with an old clustering code and I can replicate the issue. Thus some models might be missing from the irmsd file and thus the issue lies before clustering. Check if the models are created in the "rotated" folder and if they are there then check in the "merged" folder to further narrow down where the problem lies.

output_irmsd.txt

[JanAarts]

Can you add the error and output logs for these cases? @Danielle-314 I remember that you showed these models thus they are generated at least in rotated/merged?

[Danielle-314]

Cases 679, 898, 908 and 992 have a merged and rotated model. The models are also present in the irmsd file, I have received and send. I even have calculated the average irmsd for these cases, which is high (picture appended). However, the highest irmsd average in the included cases was 23.6.

Here is the error and output log: job_12595397_swifttcr_SS_tcr118_dr_m3_m0_out.txt job_12595397_swifttcr_SS_tcr118_dr_m3_m0_err.txt
I have not seen anything unusual in the error and output file.

[JanAarts]

updated_create_dict.txt

Thank you Danielle, I overlooked that. I have fixed the bug. It was caused when models are not similar to any other models within threshold then they were not initialized as an empty cluster candidate with 0 neighbors. Can you push the code into the new clustering.py (I do not have access to it on my machine). The function now has an additional else statement which includes models without neighbors. You can replace it with the old function.

See attachement