Question about running cellSNP on merged BAM files from multiple samples about cellsnp

[huangyh09]

Hi Lukas,

Thanks for follow up the discussion here. It is indeed nice to test the synthetic mixture before the real experiments.

First, I think samtoools merge can't resolve the issue that cell barcodes are overlapped between multiple samples, so the merged cell barcode may refer to multiple cells in different donors. That's the main motivation that we wrote the python simulate, which we mainly manipulated the cell barcode by adding the donor id (as the lane id), so it will be barcode-1, barcode-2, etc. If you already keep the donor id in the lane id, then I guess samtools should be fine.

For the our python simulator, it requires cell barcodes to avoide parsing all whitelist, as most are empty drops. The input barcodes files should be each barcodes for each bam. For example if you pooling 4 samples, it will be -b path/barocde_file1.tsv,path/barcode_file2.tsv,path/barocde_file3.tsv,path/barcode_file4.tsv in the same order of bam files. BTW, the barcode file need to be plain text but not gzipped (sorry developped with cellrange v2).

Once you pool the bam files properly, you run cellSNP with giving a list of SNPs and cell barcodes, it should run in mode 1.

Let me know anything unclear.

Yuanhua

[lmweber]

Ok, thank you! I will try to merge the BAM files correctly, and then get back to you again if I have any further questions.

[lmweber]

Hi Yuanhua, thanks a lot for the reply, and for providing the synth_pool.py script. This makes sense -- so the way I was trying to merge the BAM files was too simple.

I have now tried running synth_pool.py together with the Cell Ranger BAM files and (gunzipped) filtered barcode lists per sample, to merge the BAM files, but now I am getting the following error:

Traceback (most recent call last):
  File "../synth_pool/synth_pool.py", line 266, in <module>
    main()
  File "../synth_pool/synth_pool.py", line 220, in main
    load_sample=False)
  File "/users/lweber/.local/lib/python3.7/site-packages/cellSNP/utils/vcf_utils.py", line 70, in load_VCF
    if vcf_file[-3:] == ".gz" or vcf_file[-4:] == ".bgz":
TypeError: 'NoneType' object is not subscriptable

If I understand this correctly, it is because I did not provide a VCF file to the --regionFile argument. However, if I provide a VCF file, then the script will return a BAM file containing only reads covering the given variants -- but I am trying to merge the BAM files and keep all reads.

Is there a way to run the script to merge BAM files without the --regionFile argument? (Sorry I am not so proficient at Python, otherwise I could probably just edit the script.)

[huangyh09]

Hi Lukas, you understanding is correct. This code indeed requries --regionFile, as it parses reads by covering a list of SNPs (see the code). You could change this section of codes to cover the whole chromosome(s). However, this stragegy is probably not fast, and difficult to handle many settings.

Alternatively, you have the initial cell barcodes with the sample id when running cellranger (not sure if --sample or --lanes can do this?), then you can use samtools merge, which might be more efficient to pool the whole bam files.

Yuanhua