Tweet
We are analyzing bam files using bam-readcount and bcftools + mpileup, and comparing the reported level of coverage at specific sites in the genome. We compared both raw read coverage depth and the coverage depth where reads/nucleotides are filtered. Read filtration is performed using similar parameters in both programs
While we do not expect results to be identical, some regions showed markedly different coverage for filtered reads between the two programs. These regions were within repetitive regions. The repetitive regions themselves contained possible indels or misalignments. When looking at the regions in IGV, the potential indels/misalignments where not in the specific regions showing strong differences between the two programs.
One difference between the algorithms used by bcftools and bam-readcount to calculate per-site coverage is that bam-readcount will count indels (i.e. '-' symbols in column 10 in the bam/sam file) and bcftools will not. I understand that bcf-tools also performs a local realignment around repetitive regions.
Could realigning the repetitive region be responsible for this difference in difference in coverage estimates? Are the realigned regions left- or right-aligned?
An example of our coverage results is below:
Chromosome Position bcr_raw brc_filtered bcft_raw bcft_filtered
chr5 112111314 204 202 204 131
chr5 112111315 205 203 205 131
chr5 112111316 206 204 206 132
chr5 112111317 212 210 212 138
chr5 112111318 212 208 212 137
chr5 112111319 212 207 212 137
chr5 112111320 212 206 212 137
chr5 112111321 213 207 212 139
chr5 112111322 211 195 210 135
While we do not expect results to be identical, some regions showed markedly different coverage for filtered reads between the two programs. These regions were within repetitive regions. The repetitive regions themselves contained possible indels or misalignments. When looking at the regions in IGV, the potential indels/misalignments where not in the specific regions showing strong differences between the two programs.
One difference between the algorithms used by bcftools and bam-readcount to calculate per-site coverage is that bam-readcount will count indels (i.e. '-' symbols in column 10 in the bam/sam file) and bcftools will not. I understand that bcf-tools also performs a local realignment around repetitive regions.
Could realigning the repetitive region be responsible for this difference in difference in coverage estimates? Are the realigned regions left- or right-aligned?
An example of our coverage results is below:
Chromosome Position bcr_raw brc_filtered bcft_raw bcft_filtered
chr5 112111314 204 202 204 131
chr5 112111315 205 203 205 131
chr5 112111316 206 204 206 132
chr5 112111317 212 210 212 138
chr5 112111318 212 208 212 137
chr5 112111319 212 207 212 137
chr5 112111320 212 206 212 137
chr5 112111321 213 207 212 139
chr5 112111322 211 195 210 135