search

kharchenkolab / numbat

Haplotype-aware CNV analysis from single-cell RNA-seq
https://kharchenkolab.github.io/numbat/
Other
163 stars 23 forks source link

CNV calling from 10X 5 prime GEX data #193

Open cv55 opened 2 months ago

cv55 commented 2 months ago

Hello,

I have been using Numbat for quite sometime now, and it has worked great. Recently we switched to the 5' 10X protocol. Numbat fails to correctly identify the CNVs, and in all samples i keep getting an amplification on chromosomes 16 and 19 (which is prone to false positives anyway) , which is not correct.

Is this something you have been aware of? Could the 5' affect the performance of the tool so much?

teng-gao commented 2 months ago

Hello, it shouldn't make that big a difference. If the normal expression reference is from 3' and you're analyzing a 5'-sequenced tumor sample that might cause more noise.

DarioS commented 1 month ago

Also, we have had trouble with whole genome duplicated samples and are quietly hoping for a Numbat update. Is there any difference in W.G.D. rate between batches if you have matched whole-genome sequencing data to cross-check? If you look at my heatmap, all of the samples with inferred chr16 amplification by Numbat have W.G.D. and disagree with W.G.S.