Issue Running CONICSMat Tutorial Due To Deprecated SCRAN Function

[MayaTalukdar]

Hello,

I am working through the SmartSeq2 scRNA seq of Oligodendroglioma tutorial for CONICSMat and running into an issue with this part:

vg=detectVarGenes(suva_expr,500)
ts=calculateTsne(suva_expr,vg)
plotTsneGene(ts,suva_expr,c("MBP","CSF1R","ALDOC","OLIG1"))

It appears that detectVarGenes calls a scran function TrendVar, which is deprecated. SCRAN suggests replacement with fitTrendVar, but this takes in both a matrix of counts and a vector of variances, while TrendVar only takes in the count matrix. How can we generate this vector of variances in order to successfully run fitTrendVar and therefore visualize CNVs on a tSNE?

Thank you!

[UmairSeemab]

I have the same issue

[nlaillerbms]

Same problem

[Bakr-Mo]

same here

[andynkili]

same here

[jcostea]

same problem

[Aggarwal-Ayush]

Hi

If anyone's still interested in the solution for this or comes looking for it in the future, they can use the below modified function to detect variable genes.

detectVarGenes_custom = function(mat,ngenes=500,FDR=0.00001){
var.out = scran::modelGeneVar(mat)
hvg.out = var.out[which(var.out$FDR <= FDR),]
hvg.out = hvg.out[order(hvg.out$bio, decreasing=TRUE),]
genes=rownames(hvg.out)[1:ngenes]
return(genes)
}