Gistic2 Call

From Array Suite Wiki

In TCGA Land

In TCGA land, Copy Number data are provided in two types, GISTIC2 and Log2Ratio-based calling. GISTIC2 call results are downloaded from Broad Firehouse Runs. *.CopyNumber_Gistic2.Level_4.*.tar.gz files are used to get the GISTIC2 call values. GISTIC2 views are displayed in three Land views (These views are defaulted to use this data) that summarize the Copy Number status of a gene in the filtered/samples:

CNV Gistic views.png

There are six calls:

  • Amplification: 2 or more extra copies
  • Homozygous Deletion: 2 missing copies
  • Gain: 1 extra copy
  • Heterozygous Deletion: 1 missing copy
  • Copy Neutral LOH: proper copy number, but loss of heterozygosity
  • Diploid: normal copy number

By default only samples with a status of "Amplification" and "Homozygous Deletion" checked for all views with GISTIC2 data:







All three views offer the user the ability to toggle to Log2Ratio-based segmentation calling method. Details for this data can be found here.


In the details window for selected samples, you can click the small arrow on the left side of detail window and choose Copy Number Details to find Gistic call along with log2ratio values for each sample for the searched gene.


Note: Users may notice discordance between CNV calls from Gistic2 and log2 segment data. This is due to differences in the the data types and algorithms that are used to call the CNV. These differences can result in the following observations:

CNV status does not match: An example is in the image above, where the second sample listed is reported to have a Homozygous deletion (by GISTIC2), but diploid (by segment copy number status).

CNVCall present, CNV Segmentation absent: The gene coordinates do not lie within any CNV segmentation, but CNVCall is present because Gistic2 algorithm provides calls across the entire chromosome even if there are gaps in the CNV segmentation data. In this case, genes will not have any views associated with log2 segment data.

CNV Segmentation present, CNVCall absent: The CNVCall is a matrix of gene names (rows) x samples (columns). Some gene names in the CNVCall data do not completely match the gene names in our gene model and are therefore not represented in the GISTIC2 data viewed in TCGA.