varscan report wrong DP related values, almost 50% off

[yingchen69]

Hi,

I just found a wired issue with varscan and it happened to both 2.4.4 and 2.4.3. Basically the varscan vcf files have wrong DP related values, about 50% off the values reported by IGV or Mutect2 vcf files.

Here is the varscan 2.4.4 vcf for EGFR L858R:

chr7 55259515 . T G . VarBaseQual ADP=1191;WT=0;HET=1;HOM=0;NC=0;ANN=G|missense_variant|MODERATE|EGFR|EGFR|transcript|NM_005228.3|protein_coding|21/28|c.2573T>G|p.L858R|2819/5600|2573/3633|858/1210||,G|sequence_feature|LOW|EGFR|EGFR|helix:combinatorial_evidence_used_in_manual_assertion|NM_005228.3|protein_coding|21/28|c.2573T>G||||||,G|upstream_gene_variant|MODIFIER|EGFR-AS1|EGFR-AS1|transcript|NR_047551.1|pseudogene||n.-2873A>C|||||2873| GT:GQ:SDP:DP:RD:AD:FREQ:PVAL:RBQ:ABQ:RDF:RDR:ADF:ADR 0/1:27:1191:1191:1179:12:1.01%:1.6693E-3:53:54:1013:166:11:1

Here is gatk Mutect2 vcf for the same variant:

chr7 55259515 . T G . PASS AC=1;AF=0.500;AN=2;CONTQ=93;ClippingRankSum=0.125;DP=2318;ECNT=2;FS=0.000;GERMQ=93;LikelihoodRankSum=-1.190;MBQ=20,20;MFRL=172,176;MMQ=60,60;MPOS=37;MQ=60.00;MQRankSum=0.000;POPAF=7.30;ROQ=77;ReadPosRankSum=0.086;SEQQ=28;SOR=0.615;STRANDQ=31;TLOD=6.62;UNIQ_ALT_READ_COUNT=26;ANN=G|missense_variant|MODERATE|EGFR|EGFR|transcript|NM_005228.3|protein_coding|21/28|c.2573T>G|p.L858R|2819/5600|2573/3633|858/1210||,G|sequence_feature|LOW|EGFR|EGFR|helix:combinatorial_evidence_used_in_manual_assertion|NM_005228.3|protein_coding|21/28|c.2573T>G||||||,G|upstream_gene_variant|MODIFIER|EGFR-AS1|EGFR-AS1|transcript|NR_047551.1|pseudogene||n.-2873A>C|||||2873| GT:AD:AF:DP:F1R2:F2R1:SB 0/1:2146,26:9.246e-03:2172:1016,11:1122,14:1029,1117,13,13

When I look at the bam file in IGV, the DP numbers from IGV are almost identical to the values from Mutect2 vcf.

Any explanation?

Thanks,

Ying

[bioinfolusx]

I have the same question.

[nh13]

Have you take a look at the FAQ- Read counts from VarScan are different from SAMtools/IGV counts?

Also, the "DP" reported by Varscan2 seems to be across all the reads not just those that observe the called alleles? For example, if there are three candidate alleles A, C, and G, but the genotype was A/G. Then DP would be the sum of reads observing the three candidate alleles, while AD and RD would be for the G and A allele respectively (assuming A is the reference). Note that it depends on the samtools pileup, so if you've filtered bases there, then that would also reduce the depth counts.