Self Organizing Maps for RNA-seq data

Question

[pixie@bioinfo]

Hello, I have normalized reads from RNA-seq data and I am trying to build SOMs using the oposSOM R package. It exports couple of CSV files like K-means, over-expression modules etc. However, I dont see any option to export gene lists for Correlation Cluster, Underexpression Spots. Any one used this package or knows how to reach the authors ? Any other tool I may use for SOMs ? Thanks.

https://bioconductor.org/packages/release/bioc/html/oposSOM.html

Answer 1

[Mensur Dlakic]

Both ESOM and somoclu are excellent tools for SOMs, though I don't know if they will produce exactly the output you need.

I have no experience in analyzing RNA-seq data using SOMs, but the consensus seems to be that t-SNE or UMAP work better in this domain. I have used all three with other data types, and the only advantage of SOMs is that the cell spacing is more indicative of actual distances between data points. However, when it comes to actual clustering, the other two have usually produced better results in my hands.

[pixie@bioinfo]

Thanks for the suggestions. Will try this out.

Answer 2

[pixie@bioinfo]

Other than Mensur's answer which is great, I managed to get a reply from the authors. Maybe useful if anyone wants to try this out as this is not mentioned anywhere in the R documentation

modules.CSV.sheets <- oposSOM:::modules.CSV.sheets
environment(modules.CSV.sheets) <- env

modules.CSV.sheets(spot.list=env$spot.list.correlation, main="Correlation Cluster", path=file.path(env$output.paths["CSV"], "Spot Lists") )
modules.CSV.sheets(spot.list=env$spot.list.underexpression, main="Underexpression", path=file.path(env$output.paths["CSV"], "Spot Lists") )