I was trying to check SNV for some genes other than the ones presented in the paper or figures. The files I could find are In Files -> legacy -> 20190328_data_freeze -> Files -> GLASS-M2vcf.tar.gz, and there are 280 vcf files inside it. I supposed I could get SNV information from these files but not sure of it. So I checked some common SNVs like IDH1 first. According to the figure 3 or the raw data of that figure (Files -> legacy -> 20190328_data_freeze -> Files -> figure1_3_input.RData), there seem to be 86 patients with IDH1 mutations. But from these vcf files, only 14 patients have IDH1 R132 mutations. A quick example is GLSS-HF-3081, there is IDH1 R132H mutation based on the data from the variable ?hmapdata? of figure1_3_input.RData. But according to the file of GLSS-HF-3081.vcf from GLASS-M2vcf.tar.gz, there is no IDH1 mutation for this patient. Other genes like TP53 and ATRX seem to be matched. Just wondering if I am missing anything?? Thanks!

Created by Lee Chen leechen
Hi Kevin, thanks for your clarification! Sorry I did not see the previous thread or was not aware they might be the same issue. Thanks for providing this information! -Lee
Hi @leechen, You are not missing something. I'll try to provide the background on this and hope it clarifies things. It is also briefly discussed in a prior thread: https://www.synapse.org/#!Synapse:syn17038081/discussion/threadId=6921 We all know that IDH1, IDH2, and TERT promoter are frequently mutated in glioma at genomic locations prone to mutation often referred to as hotspots. Given their importance in glioma biology and subtyping, the GLASS mutation calling pipeline uses Mutect2 to "force call" somatic variants at these hotspots using the GENOTYPE_GIVEN_ALLELES argument that forces Mutect2 to genotype these variants (i.e., IDH1, IDH2, and TERTp known variants) in every sample. We previously observed that Mutect2 missed hotspot variants when using the GENOTYPE_GIVEN_ALLELES argument in the GLASS dataset where many tumor samples whose clinical tests had confirmed an IDH mutation via sensitive clinical tests did not have an IDH mutation call that had passed filters. A former member of the lab, Floris Barthel, dug into this issue and reported a bug in the Mutect2 software (https://github.com/broadinstitute/gatk/issues/5695), but it was not fixed prior to our analyses. We recovered these filtered out mutations in our internal lab database for these hotspot variants. Thus, they would be present in those figures/tables but not in the VCF files. This approach would not have impacted genes like TP53 and ATRX that were not force called. -Kevin
Yeah, that is odd, then.
Hi Vincent, Yes they could be IDH2. But in the figure 3A of the paper and in the variable "hmapdata" from the file figure1_3_input.RData, it is showing IDH1 and specifically R132 mutations for all the IDH mutations. So I supposed they are all IDH1 mutations? And I checked the vcf files to see if there are IDH2 mutations. There is only one sample with nonsynonymous SNVs and is not the hotspot mutation IDH2 usually has (R172). So... I don't know if I am missing anything else...?
lee - im not affiliated with GLASS, so please take with a grain of salt, but i had two thoughts you could check: 1) in the clinical annotation files, the it is marked as IDHmt or WT, not IDH1 or 2. Could it be the others are IDH2 mt? 2) Alternatively, could depend on the time of sample collection / tx history (ie if treatment dropped the VAF enough could look IDHwt *by* NGS) 3) probably a more sensitive assay was used for dx

SNV data and IDH1 mutations in the dataset? page is loading…