digitalcytometry / ecotyper

EcoTyper is a machine learning framework for large-scale identification of cell states and cellular ecosystems from gene expression data.
Other
184 stars 42 forks source link

[Ecotyper] Error in NMF step at Tutorial1 using user-defined profile #88

Open jpark27 opened 8 months ago

jpark27 commented 8 months ago

Dear EcoTyper team,

Hi, all! Hope you been well and thanks so much for sharing amazing tool.

I was following tutorials using new TCGA data with in-house scRNA-seq dataset. Initially, I run Tutorial5 to create my own 'discovery dataset' and successfully executed (30 cell types; 200K cells).

Then, when I use that discovery dataset on TCGA data through tutorial 1, it throws following error: and seem like other user also reported same error before (https://github.com/digitalcytometry/ecotyper/issues/39)

Rscript EcoTyper_recovery_bulk.R -d discovery_scRNA_UVM -m /lustre/scratch126/cellgen/team283/jp27/02_tools/ecotyper/tcga/bulk_uvm_data.txt -a /lustre/scratch126/cellgen/team283/jp27/02_tools/ecotyper/tcga/bulk_uvm_annotation.txt -c type -o /lustre/scratch126/cellgen/team283/jp27/02_tools/ecotyper/tcga/DiscoveryOutput_bulk_uvm

image

(*Surprisingly, this new TCGA data run without error with pre-defiend -d Carcinoma or Lymphoma profile. So I thought bulk RNA-seq data itself has no issue but on newly generated 'discovery dataset' [discovery_scRNA_UVM])

Any advices how to resolve this error? I can happy to share any intermediate/input files I generated if it's useful to troubleshoot.

best wishes, J

Yuchen588 commented 7 months ago

I've encountered the same error when using pretrained scRNA-seq data for deconvolution and estimating cell state and ecotypes in bulk tissue profiles. Could you please share your complete code through a private channel so we can examine the details? Thank you! @jpark27