search

broadinstitute / infercnv

Inferring CNV from Single-Cell RNA-Seq
Other
563 stars 166 forks source link

inferCNV can not analysis the same data like before jump out with "span is too small" #499

Open yintz opened 1 year ago

yintz commented 1 year ago

hi, when I use inferCNV new version to run the script with the same data(worked before) as control, it suddenly jumped out and stopped to say:

Error in simpleLoess(y,x,w, span degree, parametric = parametric, : span is too small

then, it jumped out from the analysis. is there any idea what happened? I run the same analysis for same data in august, it could generate the plot nicely. but now, it can not.

I also tried to degrade the inferCNV with old version like 0.99.0, 0.99,7 1.13, 1.14, 1.15, it all didn't work now even with same script. do you think this is due to other updated dependency packages and how could we fix it?

is there any preferences on the data? I noticed, If I set the expression 3 times higher, it do show more CNV status, but /3 it shows lower CNV and easily to get into span is too small issue. if we considered expression, this 3 or /3 should not make any differences in theory. Don't we?

any suggestions?

GeorgescuC commented 1 year ago

Hi @yintz ,

The initial absolute values of the expression are used to filter non expressed genes using the cutoff threshold. If you divide all expressions by 3, more genes will be filtered out, which might leave you with too few. Inversely if you multiply expression by 3, less genes will be filtered out and you might dilute signal.

Regards, Christophe.