Cell lineage identification from pathway anlaysis

Question

[rob.costa1234]

From RNA seq data I am trying to find out the cell types enriched in particular cluster from DE genes, Is there any specific pathway or tool that can allow me to assign the cell identity using from DE genes. thanks

Answer 1

[Kevin Blighe]

What you are aiming to do is generally referred to 'cellular deconvolution'. Earlier methods were designed for microarray data and it is only now that methods for bulk and single cell RNA-seq are being developed.

It is suggested that one of the earlier efforts originally developed for microarray, i.e., CIBERSORT, can be used for RNA-seq data too. Other methods are in development. A worry that I have (for bulk RNA-seq) is that it is no longer the prime method for doing what you want to do. Single cell RNA-seq is superior in order to define cellular lineages, and methods are coming out more for single cell than they are for bulk RNA-seq, in that regard.

• BSEQ-sc: Deconvolution of Bulk Sequencing Experiments using Single Cell Data
• Estimation of immune cell content in tumour tissue using single-cell RNA-seq data

You may additionally want to look at the data available by tissue-specific consortia, such as:

• FANTOM
• The Genotype-Tissue Expression (GTEx)

What I suggest is that you obtain that data, identify the genes that define each tissue type, and then 'map' (deconvolute) your own data against those custom tissue signatures. Then you will have developed your own deconvolution method.

Kevin

[rob.costa1234]

CIBERSORT has one signature genes file LM22 for blood cells. However I want to identify stem cells, mesenchymal and fibroblasts. Is there any data base where I can get signature genes (at least name of genes) for different cell types?

Thanks

[Kevin Blighe]

Those cell-types should be in the FANTOM5 dataset, however, you will have to find a custom way to overlay your data with the signatures from FANTOM5. I an give you some brief guidance, if you wan to first take a look there.

[rob.costa1234]

Kevin, Could you suggest any GUI/ web based old tool that has been used for cellular deconvolution of micro-array data? I googled it but could not find something.

[Kevin Blighe]

Is CIBERSORT not okay for that purpose? You mention it in your other comment.

[rob.costa1234]

CIBERSORT need signature genes and provide only LM22 genes as signature for immune cells. It looks lot more complex to draw our own signature from http://fantom.gsc.riken.jp/data/.

Thus in the meantime I am just trying to see how the previous cellular deconvolution of micro-array data has evolved. But I dont see much information in this regard. I think if there is a web based database where if we plug in expression matrix and it identify possible cells will be a good starting point.

[Kevin Blighe]

It hasn't been an area of intense research, put it that way, which is why there are not many options available. Another 'package' that I recall is CellMix, which allows you to use one of the most popular signatures, referred to as Abbas - this is for microarray but it is implemented in R.

I am sure that further methods will come out with the advent of GTEx and the Human Protein Atlas ( https://www.proteinatlas.org/tissue ).

You may want to take a look at this:

• Computational Deconvolution: Extracting Cell Type-Specific Information from Heterogeneous Samples

That lists some programs, and the final author of that is the guy who developed CellMix (I believe).