Feature requests to better aid histopathology diagnosis

[jjgao]

From a user:

We see diagnostically challenging histology cases for which we have extensive immunohistochemistry data but still do not know either the tumour type or the site of origin. We then perform NGS (Illumina TSO500), and /or WGS, giving us mutations, CN gain/losses and fusions. We then use cBioportal and GENIE to give clues to the site of origin or tumour type. In cbioportal we typically select the "TCGA pancancer atlas series", then enter relevant genes, sometime with specific mutation using OQL. The most useful ways for us to visualise data is the "cancer type summary" tab as this ranks possible diagnoses by their frequency, which can sometimes strongly support some diagnoses or exclude others.

This works very well for single genes, but when there are combinations of genes then the overall frequency data is not as useful as (1) the Boolean logic always appears to "OR", (2) this plot doesnt tell you which gene is affected by either mutation/CN change/SV, and (3) CN variants cannot be selected for gain vs loss. For example, if we have detected an activating BRAF mutation and homozygous NF1 deletion then it would great to be able to see the frequency plots across cancer types for BRAF GoF AND NF1 LoF, and even better, if we could specifically toggle between combinations like (BRAF V600E AND homozygous deletion of NF1); (any gain of function mutation in BRAF AND any loss of function of NF1 [e.g. including combined missense/fs truncating+CN loss]) etc. When mutation between >2 genes are involved, then toggling between various logical combinations would be very powerful (e.g. GoF BRAF AND LoF NF1 AND LoF PTEN; GoF BRAF AND [LoF NF1 OR LoF PTEN]; [GoF BRAF OR LoF NF1] OR LoF PTEN).

I understand the oncoprint shows this data with detail down to individual cases and specific variants/CN change, and the frequency of various combinations can be inferred from this, but the actual numbers of cases within specific cancer types is not readily visualised in this tab. Clicking on either study in cBioportal or GENIE give detailed summary data, and frequencies of various combination of specific mutated gene or gene with CNA can be obtained, but this is not particularly easy for histopathologists to do, but I dont think that specific mutations (e.g. BRAF V600E, or only CN loss, not CN gain) can be selected.

I know that downloading the data would likely help, but unfortunately is beyond the skill of most histopathologists. I have found the graphs in cBioportal are very powerful diagnostic aids, and any way to be able to present data for specific combinations of genes and mutations in visual format would be gratefully received.

Other improvements could also include:

• show examples on the data entry page how to enter missense mutations not just hyperlink to OQL (show e.g. BRAF:MUT= V600E)
• default to excluding VUS (these are best excluded for histopathologists as they may be unaware that these are being included)
• show visually on screen number of cases within a series or flag those with low numbers as they may not be representative.
• if the data exists, allow viewing within a cancer subtype by histologically grade (as high grade tumour can have different genomics to low grade)
• allow selection of cases to include by gender and age
• interface with immunohistochemistry profile (this may be too hard, but ultimately this is what histopathologists do with the NGS, so would be very powerful)

@cBioPortal/product

[jjgao]

The main feature request is AND logic, which can be addressed by https://github.com/cBioPortal/cbioportal/issues/9596.

[jjgao]

Another user request on AND logic: https://groups.google.com/g/cbioportal/c/2hROmV0fg9o

[jjgao]

This feature may be useful as a workaround: #9790 - selecting co-altered samples and view them in study view.