I downloaded expression (microarray), clinical and mutation data associated with METABRIC breast cancer study from cBIOPORTAL and I'm planning to analyze the dataset for DEGs among three subsets of individuals, subsetted based on a mutational status. I'm not familiar with using median and z-score expression values for downstream analysis. I was initially planning to set up a SummarizedExperiment object and use DESeq2 to analyze the data. Since these expression values are already normalized, I believe can't go down that path. Can I just run ANOVA (Kruskal-Wallis test) on expression values for each gene from three samples/subsets?
Any help on understanding what these median and z-score expression files exactly are, and how I may proceed with the analysis is much appreciated.
1 answer
If you are downloading the Z-score expression values, then —yes— you can do parametric / non-parametric tests on these, such as ANOVA / Kruskal-Wallis ANOVA. You could feasibly use limma, too, as it just fits a weighted linear regression model to your data.
I would not use DESeq2, as it would expect raw counts as input. Even if you coerce your data to a DESeq2 object, I believe that the model assumptions employed by DESeq2 for your data will still be incorrect.
Kevin
I am not sure - have you searched?
Log in to answer this question.
-
How to retrieve RNA expression dataset of METABRIC - 144 adjacent normal samples ?
written by pooja •
-
Proteomics data analysis
written by Ariadna •
-
Help doing differential expression analysis -experimental design and gProfiler TF interpretation-
written by Camilo Andres •
-
How to perform differential expression analysis of metabric data obtained from METABRIC?
written by DareDevil
-
How to sort the TCGA dataset based on expression of a panel of genes?
written by dr.goutham •
-
Nonparametric statistical tests on data with negative and positive values
written by SB •
-
cBioPortal Z-score: value difference between download and web Portal
written by faswad •
-
cBioportal survival analysis help
written by mfangli •
-
How can I present gene expression values for multiple genes in various cancers?
written by arman s •
-
Mutual exclusivity analysis on expression data
written by orzech_mag
Thanks, Kevin. I checked the distribution of data for each test category and they looked normally distributed.I did ANOVA and Kruskal-Wallis ANOVA on z-scores. Just wanted to make sure that I didn’t missing any. Also, thanks again for your answer on implementing K-W and parametric ANOVA in R. It was very helpful.
Sure thing.
Any idea on what the median data file that comes along with the download? Thanks!
Can you clarify what the file-name is?
It says data_expression_median.txt and a representative summary of data distribution for a sample looks like below;
Min. 1st Qu. Median Mean 3rd Qu. Max. 4.713 5.401 5.659 6.423 7.095 14.464
whereas for the same sample in the z-score file (data_mRNA_median_Zscores.txt) summary looks like below. Min. 1st Qu. Median Mean 3rd Qu. Max. NA's -4.1048 -0.6847 -0.0855 -0.0321 0.6074 11.4517 1006