CellMix compare results to CellCODE
Entering edit mode
Last seen 5.7 years ago

I was hoping I could get advice in terms of experimental design on a project that I am doing with CellMix.

I am studying disease WHOLE BLOOD tissue samples and I would like to get an estimate of the proportions of cells in the tissue samples and from this I need to eliminate any confounding in terms of differential expression that may be due to the cell types.

Illumina was the tech used to assay the samples and RSN (Robust spline) was used to normalise the data.

Originally this was done using CellCODE. I produced a heat map that estimated differentially expressed genes with respect to cell type. And then plotted the estimated proportions against the physically calculated cell counts to see how they compared (I have data for both of these expression and physical cell counts). This seems to be exactly what your CellMix vignette describes in the Figure 1 from the CellMix original paper: http://bioinformatics.oxfordjournals.org/content/29/17/2211.long

However the CellCODE package is still under development and CellMix seems to have a richer array of commands so I was hoping to compare them.

My problem is similar to the one that has already been asked on the Bioconductor Forum: CellMix package use with own data matrix

Using a GEO dataset as a contrast (in CellCODE I used "IRIS"), I would like to use my own expression data to estimate a the cell proportions of my tissue samples.

Based on the advice at the following link: CellMix package use with own data matrix, I have tried the following (and not got very far):



Beyond this implementing "eset" in any of the CellMix commands does not work as it is not compatible with "gedBlood()".

gedBlood(eset, verbose = TRUE)


Using ged algorithm: “lsfit”

  Estimating cell proportions from cell-specific signatures [lsfit: ls]

Error in while (any(a < 0) && length(i) < r) { : 

  missing value where TRUE/FALSE needed

The "eset" matrix is a 47323 row by 656 column dataset the rows are genes and the columns specify disease status.

I tried it with gedProportions:

> gedProportions(as.matrix(counts1), ExpressionSet(as.matrix(E.rna)))

Error in gedProportions(as.matrix(counts1), ExpressionSet(as.matrix(E.rna))) : 

  Empty data after limiting to common features: signatures [0] - target data [0]

I think that the file that contains the physical cell-counts data is incompatible with "ged"...

I should note that the file has been subject to rearrangement and has a lot of missing values. I plan to clean this data but for now I am just looking fro some preliminary instructions in terms of my experimental design.

Overall my question is how do I implement CellCODE to implement a heat map and plot similar to the ones that I plotted in CellCODE. Am I using it the right way? Can someone point me in the right direction?

cellmix cellcode • 1.4k views
Entering edit mode
Last seen 1 day ago

Both the CellMix and the CellCODE packages seem like they would be quite useful, however one problem is that they are not bioconductor packages. You might try emailing the authors directly.

That having been said, I wonder if your immediate problem isn't due to just incompatible rownames (gene identifiers). You say you are using illumina data, and I wonder if the data used for the deconvolution is from affy arrays?



Login before adding your answer.

Traffic: 297 users visited in the last hour
Help About
Access RSS

Use of this site constitutes acceptance of our User Agreement and Privacy Policy.

Powered by the version 2.3.6