@Huimin Lyu I am very sure it is different samples, they are however from the same subject, but different timepoints, not sure if this changes things. But now looking further into it, I came across something weird and it might stem from this as well. I went through all the results in the folder

vdj_annotation

, everything looks fine, however from folder

05-select-junction-mod3

to

06-annotate-metadata

5 out of 36 samples disappear. I checked their result from

*_jmod3_command_log.txt

and all have FAIL>0. I now went through the original metadata file, but can't seem to find something, which makes them different from the other samples. It is always just one sample from one subject, but different timepoints. But there are also subjects for which no samples are disappearing. Maybe you have any insights on what I should check here?

Hi Carlotta, I’d just like to clarify a few things so we can help you better: • could you share the command that you used to run the analysis and the configuration file. • The single-cell BCR contamination is output in the single-cell QC report. Can you share the part of the report where you see this duplicated reads? Like which table do you refer to in this report. In the single-cell QC report you see what is the percent contamination of sequences that have the exact same sequence and barcode. For this step, we do not make the barcodes unique, since we actually want to detect sequence contamination across different samples. Even if these are from the same patient, the chances that an exact same BCR sequence gets the same barcode_id is very low (even when the barcode_ids are re-used) so we consider these to be contaminants across samples.

Hi @Gisela Gabernet, I am happy to clarify. In my last message, I addressed a second issue, but both issues originate from the same run. Below is the order in which the issues arise in the pipeline: First issue: I lost five samples in folder

vdj_annotation

between the steps

05-select-junction-mod3

and

06-annotate-metadata

, and I am not sure why. All of these samples pass the test of having a nucleotide junction length that is a multiple of 3. Therefore, I am wondering what could be going wrong during the metadata annotation step that causes the samples to disappear. I checked the metadata file but could not find anything that makes these samples different from the others. I have now rerun the pipeline using a new metadata file that only includes the necessary columns and no additional information. With this setup, no samples disappear. I understood the documentation to mean that additional columns can be added if desired—is this not the case? For now, I will keep the shorter metadata file and add the additional information later during the detailed analysis of the results. Second issue: Contamination You are right—it is very unlikely that the same sequence would end up with the same barcode in two different samples. I initially misunderstood how this step works and thought this might be related to the metadata issue. However, after rechecking the Cell Ranger outputs, it appears that some samples were incorrectly assigned. I am currently rerunning this step, and the contamination error should be resolved afterward. I apologize for the confusion. So for now, only my metadata question remains: can I add additional columns? Thank you and sorry for the confusion!

Hi, Carlotta. I am happy both of your issues were solved. Additional columns are allowed in the metadata. I am surprised that the additional columns in your metadata file caused the loss of samples in "06-annotate-metadata" step. Could you please give me the two metadata files you used? I'd like to compare and check them to figure out what happened.

Hi @Huimin Lyu, I can't send it directly as privacy laws prohibit me to share all information, but I will randomize and pseudomize more and send it to you. What is your email?