Different number of variants between single cell and bulk
KEYS248 opened this issue · comments
At line 375 of the functions.R script, I get the following error
Error in `$<-.data.frame`(`*tmp*`, "id", value = c(".", ".", "rs371110082", :
replacement has 218653 rows, data has 324935
Calls: compare -> $<- -> $<-.data.frame
Execution halted
It seems that the site.frame variable gets populated by the single.cell.vcf.info variable and then attempts to fill more columns of site.frame with the bulk.vcf.info variable. The error seems to occur because bulk.vcf.info has fewer rows than single.cell.vcf.info. Is this situation not supposed to occur? Did I make a mistake with some previous step?
If I kept the rows of bulk.vcf.info that match site.frame, there are still rows of site.frame that would be empty and I'm not sure how the rest of the program will handle that.
Hi, I met the same bug at line 375 of the functions.R script. Have you solved the problem yet？
@red-t Unfortunately I did not solve it. My research group decided to move forward without using LiRA
Can you confirm that the input single-cell and bulk samples were run on the same multisample VCF file?
Can you confirm that the input single-cell and bulk samples were run on the same multisample VCF file?
Sorry, I'm not very clear about "the same multisample VCF file". As a try, I just run GATK for single-cell and bulk sample(one sample for each kind) respectively. Then run LiRA step by step for each sample. The following shows the configure files for single-cell and bulk:
single-cell:
name SRR475137
analysis_path SRR475137
reference_file human_g1k_v37.fasta
bam SRR475137.bam
vcf SRR475137.vcf.gz
gender female
sample SC
bulk F
reference_identifier hg19
phasing_software eagle
only_chromosomes 21,22
bulk:
name SRR475185
analysis_path SRR475185
reference_file human_g1k_v37.fasta
bam SRR475185.bam
vcf SRR475185.vcf.gz
gender female
sample BULK
bulk T
reference_identifier hg19
phasing_software eagle
only_chromosomes 21,22
I see that the VCF field in both is different. So I think it is likely that what I suggested is the issue (it will be unless the VCFs in both configs describe the same set of sites).
"The same multisample VCF file" refers to a VCF file with multiple sample columns describing evidence for the listed variants across multiple BAM files. It can be created using GATK. See, e.g.: https://gatk.broadinstitute.org/hc/en-us/articles/360035889971
The instructions relating to calling on multiple input GVCFs, e.g. "If you have GVCFs from multiple samples..." produce the expected LiRA input, where the multiple samples include at least the bulk and single-cell.
Given this has caused some confusion, we will update the README to clarify.
I think what @cbohrson is saying (correct me otherwise) is that the program may be expecting all samples, both single cell and bulk, to come from a single multisample VCF, instead of two VCFs (one for single cell, one for bulk). If that is the case, I believe I was not doing this so I wonder if that was the cause of my error.
OK, I just ran GATK with single sample mode before. I'll run GATK again to create a multisample VCF file, then try LiRA again. Thanks a lot!