Analysing gene expression across clinical parameters

Question

[elizabethR]

Hi

I have been doing some bioinformatic analysis of TCGA data as an adjunct to my PhD. I am wanting to take this a little further and analyse RNASeq expression data according to clinical parameters also, such as disease stage, age and sex, histological markers of invasion etc. Has anyone done this analysis? What sort of statistical analyses do you use for this? I am assuming that performing ANOVA analysis is not appropriate with such a large dataset with so many multiple comparisons being made across the dataset? Bioinformatic statistics is a very new area to me.

Also could anyone recommend packages to do this type of analysis? I have started using the amazing R studio and TCGAbiolinks but there doesnt appear to be a package in the Bioconductor guide that is suitable for this.

Really grateful for any advice guys :-)

[Shamim Sarhadi]

Hi elizabethR I am not very familiar with TCGA data but If you want to do a class comparison test between two or more phenotypes ,first you should preprocess your data(including log transformation,summarization,normalization),if you want to use GEO data,I offer you using InsilicoDB https://insilicodb.com to retrieve preprocessed data, and it provides you a pipeline to do your job simply.In addition limma is a robust package for analyzing gene expression data it could create a liner model for finding markers of each phenotype https://bioconductor.org/packages/release/bioc/html/limma.html

Answer 1

[elizabethR]

Thank you guys, EagleEye that link was very useful. EdgeR manual says you should use raw counts that haven't been normalised because it's normalised and log transformed as part of its mathematical modelling. I've used edgeR to perform differential expression analysis across clinical parameters

[EagleEye]

In that case have a look at this post,

A: How to work with Level 3 data (RPKM values) from TCGA database

[lawarde.ankita1]

Hi elizabethR,

I am working on TCGA data of Lung cancer adenocarcinoma. Refereeing to your above question and the comments, as you said that you have used edgeR to perform differential expression analysis across clinical parameters, me also working on the same type of analysis. I want to find the DEG between tumor stages (Stage I to IV) but im not able to find the right way to do so. here is the my post on biostars where i have provided the R code i have used and how i have got the samples from the FireBrowse data.

Not able to get DEG from FireBrowse data using R

I think i'm going wrong at some posint in the analysis, If you have a sample R code, please share it with me or please help me find what is wrong I am doing in the DEG analysis.

Thank you in advance

Answer 2

[EagleEye]

Check this out, A: Use of TCGA database to get information on protein Expression