We ran into some GATK4_acnv failures while processing StandUp2Cancer exomes this weekend using
We've run this version of GATK4_acnv on hundreds of exomes without failing but this weekend three samples had failures. There were two types of failure messages. Two samples failed with stderr message:
java.lang.IllegalArgumentException: Must have a positive number of SNPs to perform SNP segmentation.
and one failure was a stderr message:
java.lang.IllegalArgumentException: additional variance must be positive.
Would a more recent gatk jar fix these issues? We are somewhat hesitant to
try new versions because we've done quite a bit of analysis already with this version and a switch of versions could trigger another round of work (i.e. ABSOLUTE reviewing). The common het sites were input as gs://firecloud-tcga-open-access/tutorial/reference/only_very_good_hets.tsv.
P.S. These hg19 bams were produced at Broad and seem to have reasonably decent coverage (~100x mean target coverage).