WHAMG call more DUPs in several samples compared to other software

[Linaria219]

Hi, I am using WHAMG to perform CNV analysis on individuals of a population.
Using code:
whamg -f BAM -a GRCh38.fa -e ${list} -x 10 2> err_file > vcf

After calling for each individual, the following filter was used:

1. Removes CNVs less than 50bp or greater than 1Mb.
2. Removes DELs that don't have "TF" greater than 3.
3. Removes DUPs that do not have "U" greater than 3.
4. Removes SVs that have a max(CW)< 0.2.

After completing the filter, I found that there are about 60 (~3%) individuals with more DUPs than other individuals. For the other individuals, the average number of DUPs called in each individual is 280, but for these about 60 samples, 1000 to 410,000 DUPs can be called. Further inspection of the DUPs in these 60 individuals revealed that:

1. In vcf, the CW scores of DUP and INV in the DUPs that are not shared between these individuals and other individuals are almost the same, and both scores are less than or equal to 0.5.
2. These unshared DUPs are distributed in almost every region of the genome.
3. Using IGV to view the BAM of these individuals, it was found that these individuals contained more split reads and discordant reads.

4. In LUMPY, the number of DUPs in these individuals did not show abnormality.

In many previous articles using WHAMG, this situation did not occur in the call set, especially the order of magnitude difference in the number of SVs. I would like to ask you, do you think this situation is more likely to be caused by a problem with my calling, or is it caused by the characteristics of these BAMs themselves?

[zeeev]

Hi @Linaria219,

Does the increase in duplication calls in these samples correlate with depth?

[Linaria219]

Hi, I used the following commands to calculate the average read depth of the entire genome, the average read depth of the singleton DUP region, and the average read depth of the shared DUP:
samtools depth -a BAM > depth
awk '$3>0 {s+=$3;n++}END{print s/n}' depth > mean.depth

samtools depth -a -b singleton.bed BAM > singleton.depth
awk '$3>0 {s+=$3;n++}END{print s/n}' singleton.depth > mean.singleton.depth

samtools depth -a -b sharedDUP.bed BAM > sharedDUP.depth
awk '$3>0 {s+=$3;n++}END{print s/n}' sharedDUP.depth > mean.sharedDUP.depth

For one of the individuals with a large number of DUPs(400,524), the read depth is as follows:
mean.depth: 43.5813
mean.singleton.depth: 67.4616
mean.sharedDUP.depth: 58.9435

For one of the individuals with a normal number of DUPs(487), the read depth is as follows:
mean.depth: 44.5566
mean.singleton.depth: 50.9133
mean.sharedDUP.depth: 79.0299

It seems that the number of DUPs in individuals with a large number of DUPs is not positively correlated with read depth.

#########################################################################################
I use the preprocessing provided by LUMPY to extract the alignments:

Extract the discordant paired-end alignments.
samtools view -b -F 1294 sample.bam > sample.discordants.unsorted.bam

Extract the split-read alignments
samtools view -h sample.bam
| scripts/extractSplitReads_BwaMem -i stdin
| samtools view -Sb -
> sample.splitters.unsorted.bam

In individuals with a large number of DUPs, we can find split reads and discordant reads that are much larger than those in normal individuals. For example, in chr1:197,737,000-197,765,000, the above command can extract 3,137 split reads and 261 discordant reads from abnormal individuals, while normal individuals can only extract 22 split reads and 54 discordant reads. In this area, lumpy did not find SV, but whamg still had 26 DUPs after filtering. Can this information help answer the question?