novartis / granulator Goto Github PK

HI I am doing the following.

decon <- deconvolute(m = tata, sigMatrix = sigList)

and getting following erroe

Error in order(colnames(x)) : argument 1 is not a vector

and my data is

str(tata)
int [1:55365, 1:60] 0 0 212 16 35 807 12 0 447 1 ...

• attr(*, "dimnames")=List of 2
  ..$ : chr [1:55365] "0610005C13Rik" "0610006L08Rik" "0610009B22Rik" "0610009E02Rik" ...
  ..$ : chr [1:60] "GC.SL.625_1" "GC.SL.625_2" "GC.SL.625_3" "GC.SL.625_4" ...

Dear Authors ,

Thank you for reading my message.

I was trying to run deconvolution on my Bulk RNA seq data with multiple samples and I found your to be convenient as I had problems deciding on which method to use. I have my sample matrix as normalized TPMs and I obtained gene signatures of cell types from "celldex" and transformed the provide log counts (normalized as TPM) using :
TPMs_count = exp(log_counts)
and I ran this
> decon_rls <- deconvolute(m = TomEMD_dec, sigMatrix = cell_types_means_m, "rls", use_cores= 1)
and obtained this error only running rls

Running deconvolution method "rls" on signature matrix "sig1"
Error in rlm.default(df$x, as.vector(z), maxit = 1000, psi = psi.huber) :
'x' is singular: singular fits are not implemented in 'rlm'

and this error when running dtangle
> decon_dtangle <- deconvolute(m = TomEMD_dec, sigMatrix = cell_types_means_m, "dtangle", use_cores= 1)

Running deconvolution method "dtangle" on signature matrix "sig1"
Warning message:
In max(rowSums(x), na.rm = TRUE) :
no non-missing arguments to max; returning -Inf

and when running svr it run nonstop without finishing

other methods like 'ols','nnls','qprog' run without any problem

Could you help me ?

Thanks,
Best

Hello, as in the other issue, how can we generate a reference profile starting from RNAseq data?
What I did was to use one single cell type from the raw reads counts to generate the reference but I have doubts about the low deepness. Should I sum the counts of all the cells from the scRNAseq ( i.e. cells of the same type)? or we need to use a mathematical model (regression..) to obtain the couns starting from the scRNAseq data? Could you explain how did you obtain your reference profile files?
Thanks.

Dear granulator_team,

Thank you for providing this fantastic tool.
In the tutorial, you suggest using a ground_truth file to benchmark the various deconvolution methods. How should this "ground_truth" be retrieved in practise, e.g. what kind of data would constitute the ground truth if I have a bulk expression dataset from a solid tumor?

Thanks ever so much.

Best wishes
Othman

Thanks for providing this interesting tool. Cool you explain more in-depth how to generate reference profiles? In the vignette you state that it
"... consists of a gene (rows) by cell type (columns) matrix containing transcript-per-million (TPM) gene expression values normalized for total mRNA abundance."

Further down, you write that
"Reference profiles for deconvolution are usually generated by differential expression analysis on bulk RNA-seq generated from isolated cell types or cell-type clusters identified by single cell RNA-seq."

What measure would you want me to use from DE analysis on scRNA-seq clusters as reference? Also, is this DE analysis a comparison between clusters or between conditions? And how would you normalize for total mRNA abundance from scRNA-seq data?

Optimally, some examples for creating reference profiles would make it really easy for people to start using your tool.

BR

Rasmus