we attempted to replicate brass results using the implementation in analysisWGS.sh on pcawg samples, but we are unfortunately getting very different results for some of the samples. Take for example DO45191_tumor, where pcawg's brass vcf reports only 270 rows in the vcf, but ours has >45000 rows, and the pcawg calls seem to be a subset of our calls. Interestingly, only 2 of 42 pcawg samples had such a dramatic difference between them, although for most of the samples, our calls were supersets of the pcawg-generated calls. as far as i can tell i used the implementation of analysisWGS.sh pretty much to the letter, even separating out the input and cover steps from the downstream.
We are now just wondering if there's any extra filtering steps to perform on brass to pare down that insanely high number, or if you would be able to replicate our results on your end using the same sample. i am currently using the brass version 6.3.4
we attempted to replicate brass results using the implementation in
analysisWGS.shon pcawg samples, but we are unfortunately getting very different results for some of the samples. Take for exampleDO45191_tumor, where pcawg's brass vcf reports only 270 rows in the vcf, but ours has >45000 rows, and the pcawg calls seem to be a subset of our calls. Interestingly, only 2 of 42 pcawg samples had such a dramatic difference between them, although for most of the samples, our calls were supersets of the pcawg-generated calls. as far as i can tell i used the implementation ofanalysisWGS.shpretty much to the letter, even separating out theinputandcoversteps from the downstream.We are now just wondering if there's any extra filtering steps to perform on brass to pare down that insanely high number, or if you would be able to replicate our results on your end using the same sample. i am currently using the brass version 6.3.4