search

aertslab / GENIE3

GENIE3 (GEne Network Inference with Ensemble of trees) R-package
26 stars 9 forks source link

unable to build a weight matrix on GENIE3 #5

Closed juugii closed 3 years ago

juugii commented 6 years ago

Hello,

I have troubles in building a weightMatrix from GENIE3:

weightMatrix <- GENIE3(exprMatrix_filtered, regulators=inputTFs, nCores=36) Error in weightMatrix[regulatorNames, ] <- weightMatrix.reg : the number of object to replace is not a multiple of the replacement size

I am following the guide from SCENIC from a scRNAseq matrix (about 12000x4000 starting matrix) step-by-step, with:

dim(exprMatrix_filtered) [1] 8467 3195 length(inputTFs) [1] 786 table(inputTFs %in% rownames(exprMatrix_filtered)) TRUE 786

What objects could have any wrong dimensions or properties ? Thank you for your help,

Best regards.

Session info ------------------------------------------------------------------ setting value
version R version 3.4.4 (2018-03-15) system x86_64, linux-gnu

GENIE3 1.0.0 2018-04-18 SCENIC 0.1.7 2018-04-19 AUCell 0.99.5 2018-04-19 RcisTarget.mm9.motifDatabases.20k 0.1.1 2018-04-18

orionzhou commented 6 years ago

That error message typically occurs when there's not enough memory on the server - just try to request more memory (I'd go at least 200GB for that size of a matrix)

hummuscience commented 4 years ago

I have a similar problem, but I am limited with the amount of memory I have (60Gb).

I have a 13915x4391 matrix to run GENIE3 on. Is there a way to make it work with the 60Gb of memory that I have? I tried reducing the number of cores used. Started with 30 which gave that error, then I went down to 20 but eventually got the same error.

Any ideas?