One gene association with cancer types

Question

[kanwarjag]

I want to analyze one gene across three cancer types - bladder, breast and colon to find out its role in cancer prognosis if any. Is there any online resource/ database is available which I can use for this purpose. I found http://bioinformatica.mty.itesm.mx:8080/Biomatec/SurvivaX.jsp but it is hard to predict and cannot select more than one datasets.

Thanks

[Denise CS]

You may want to explore the Platform within Open Targets and find out if your gene of interest is associated with those cancer types. In (the current) release 1.2 of our platform, we've got 257 diseases associated with BRCA2. You can filter those results by 'Data types' (e.g. somatic mutations) and 'Therapeutic areas' (e.g. neoplasm). We are looking into adding new filters, so if you have a wish list, let us know. We may be able to implement them, and 'cancer types' seem a valid one.

Answer 1

[Alex Reynolds]

The NIH Cancer Genome Atlas (TCGA) has various data for 20.5k genes and 31 cancer tissue types, with normal and tumor subtypes for most tissues. I wrote a browser for normalized per-gene expression data which allows at-a-glance comparison of relative RNA abundance in selected tissues and either a "single" or "gallery" listing of a gene- or genes-of-interest, resp.

[kanwarjag]

Great tool. Very helpful. One question in certain cases we cannot select one type let us say for bladder only normal can be selected. Is is that tumor data can not be imported by tool or I am missing something?

Thanks

[Alex Reynolds]

Some tissue types in TCGA only offered RNA-seq data for either normal or tumor. Most offered both, however, so that is what I imported.

[Alex Reynolds]

Bladder tumor data are now available. Thanks for the feedback!

[svlachavas]

Dear Alex, just a naive question about your wonderful browser above:

I have knowledge that RNA-seq and microarrays have some major differences in experimental design-but can I use your above tool, for instance if I'm aware of a specific tumor comparison i have performed in microarrays and extracted DE genes(i.e. colorectal cancer-also the majority of the tumors are primary adenocarcinomas), and check for some selected genes of high importance, if they are also show "concordance" in "alteration of expression" in the browser, just for further "validation"-inspection of my results? For instance, if a specific gene shows also "downregulation" (maybe more accurate definition is lower RNA abundance) in the specific type of cancer?

Best,
Efstathios

[Alex Reynolds]

It might offer an extra datapoint. Notched box plots draw a 95% confidence interval around the median abundance level. If the CIs overlap between tumor and normal, then this may support the null hypothesis of no significant difference between their medians, while no overlap may support the alternative hypothesis that their medians differ.

[svlachavas]

Dear Alex,

thank you for your answer and also the very crucial point you highlighted: thus, I pick one example just for illustration - I used the gene symbol FGFR2 - which I have identified downregulated in my colon cancer vs control samples (log fold change ~ -1.3-downregulated) - So, with this pic "agrees" with my finding for this specific gene, right?https://tools.stamlab.org/tcga/?gene=FGFR2 (taking into account also the notion about the median abundance level)

[Alex Reynolds]

Turning off all the tissues, except the colon tumor and normal, shows notches that do not overlap, which suggests that the median abundances of FGFR2 RNA do not significantly overlap between tumor and normal conditions.

[svlachavas]

Dear Alex,

please excuse me for questioning again: so I would not turn off the other tissues to visualize only the one I want to compare? Or it does not oppose a problem? And if my approach is correct, then from the above plot, I can "naively" assume that there is a significant difference between the medians in the two groups? (and also with the group median of the normal being higher-which maybe "also" naively explain also the similar down-regulation of my above gene-as the comparisons I made is cancer vs control samples)?

[Alex Reynolds]

Sorry: I'm just turning off all the other tissues so that I can focus on the tissue and conditions I want. It doesn't change the data, it just makes things less cluttered.

[svlachavas]

Ok i got it !! So my approach is correct. Thank you again for your consideration on this matter !!

[Alex Reynolds]

You're welcome!

[kanwarjag]

removing message as I solved the question myself

Answer 2

[kanwarjag]

I have few clarifications about this TCGA expression browser 1. It takes level 3 TCGA data or some other level. 2. The figure shows TPM, I was wondering how the RNA seq data has bee finally analyzed in the backend. IS it RSEM or RPKM level analysis 3. Raw CSV will output raw read count data or it has been preprocessed in some way.

[WouterDeCoster]

It's probably a better idea to create a separate topic for this question.