Inconsistencies between studies

[Laura Parrilla Monge]

Hi,

I have noticed that two of the liver cancer studies (TCGA Firehose Legacy and TCGA GDC 2025) have the exact same patients, but the information about their mutations and copy number alterations (CNA) is very different. 

I have been told the discrepancy lies in that the Firehose Legacy study was processed using GISTIC and the GDC study was processed with ASCAT. I am not sure what this difference implies. Can anybody elaborate?

Can anybody give me advice on which study to use for my analysis? What I'd like to show are survival graphs and RNAseq heatmaps comparing patients, grouping them by having (or lacking) mutations and CNA in certain genes. 

Please help.

Thank you in advance,

Laura Parrilla

[Baby Anusha Satravada]

Hi Laura,

Thank you for your question.

The discrepancy is primarily due to differences in the copy number analysis pipelines used by the two studies. TCGA Firehose Legacy uses GISTIC-based copy number calls, whereas TCGA GDC 2025 uses ASCAT-based processing. As a result, the same genomic event may be classified differently between the studies (for example, as a homozygous deletion in one dataset and a shallow deletion in another).

For most downstream analyses, we generally recommend using the PanCancer Atlas cohort when available, as it includes a broader range of data types and underwent additional manual review as part of the PanCancer Atlas project. It is also important to note that the GDC cohort is based on the hg38 reference genome, while PanCancer Atlas and most other cBioPortal studies use hg19.

To ensure consistency and avoid interpretation issues, we recommend selecting a single cohort and using it consistently for mutation, copy number, expression, and survival analyses.

I hope this helps clarify the differences. Please let us know if you have any further questions.

Best regards,
Anusha