search

kieranrcampbell / tcga-gene-dosage

A Nextflow pipeline for identifying gene dosage effects from TCGA
0 stars 0 forks source link

Dosage compensation methods #1

Open mschubert opened 6 years ago

mschubert commented 6 years ago

Hi Kieran,

I hope you're doing well.

I'm wondering a bit what your plans are here and with gene dosage in general, because I've been working along similar lines in the last couple of months. Basically, quantifying to what extent genes are compensated and hyperactivated with copy number changes.

For this, I've found a 2015 paper that argued that global compensation does not take place (Fehrman et al., 2015). However, there are DE genes that mirror activation of driver mutations (Buccitelli et. al, 2017; mainly cell cycle and immune-related genes), where cell cycle also holds true when accounting for immune cell invasion (Taylor et. al, 2018).

This is all in line with our findings so far, and I'm working on a better library size and immune component estimation, to be used for TF activity and correlation network inference.

Let me know if you're interested in talking more about this!

(I'm also curious if you're ditching Snakemake for Nextflow :smiling_imp:)

kieranrcampbell commented 6 years ago

Hi Michael,

Good to hear from you!

This repo is actually an accessory to a different project (https://github.com/kieranrcampbell/clonealign) that attempts to match single-cells measured in RNA-seq space to clones in DNA-space, assuming a CNV-expression relationship. The idea was to identify a set of genes per cancer type that show a copy-number expression relationship using TCGA as a "recommended" input to the model. The paper you linked to (Fehrman et al) is really useful as it implies this is the case for a majority of genes - I think this is true even more so at the single-cell level.

Definitely interested in discussing more - perhaps email or skype? We have some interesting questions, e.g. what does the average dose-response curve look like, etc.

Can't believe someone spotted my nextflowing...still trying to decide, they both have pretty appealing features...whats your current poison of choice?

mschubert commented 6 years ago

Oh, cool! - then you might also be interested in @JEFworks' HoneyBADGER or detecting CNAs in scRNA-seq from the Marioni Lab.

I've got a fit of expected expression changes with CNAs given euploid expression and tumor purity - which could fit your TCGA expectation input. Happy to email/skype next week! (this one is quite full already)

It took me a while but I've converted to Snakemake. - Named wildcards and subworkflows are so nice, and I'd rather depend on Python than Java.