search

broadinstitute / infercnv

Inferring CNV from Single-Cell RNA-Seq
Other
557 stars 164 forks source link

Identification of malignant cells #447

Open WangJingwen21 opened 2 years ago

WangJingwen21 commented 2 years ago

Many study use InferCNV to identify malignant and normal-like cells of Epithelial. Can I use InferCNV to define malignant cells in MSS rectal cancer？

GeorgescuC commented 2 years ago

Hi @WangJingwen21 ,

It is best to have a set of cells you know are healthy to start with and use as a reference, but you should then be able to identify other cells in your observations that have CNVs as malignant. One limitation of infercnv however is that malignant cells that have no CNVs can not be detected, and at this time, copy neutral CNVs (loss of heterogeneity) or whole genome duplication are not detectable as they can not visible from expression data alone. We are currently working on an update to detect those, so it should be doable in the near future.

Regards, Christophe.

WangJingwen21 commented 1 year ago

Thank you for putting together such a comprehensive tool for use! I use normal Epithelial as reference and Epithelial in tumor as experimental data. and I got:

image image

Is it normal that no chromatic alteration in my results? which clusters of my epithelial should be consider as CNV altered?

GeorgescuC commented 1 year ago

Hi @WangJingwen21 ,

The small stretches (in number of genes) with signal also appear to occasionally appear in your small numbers of your normal-like cells, so if those are indeed all normal, I would be careful with those signals. The signal on chromosome 19 could be something of interest but it is still rather light. Out of curiosity, what filtering threshold did you use and what type of sequencing has this data been generated with? Also, how many genes are kept during the analysis?

As for the HMM predictions, it looks like the subclustering of references is too fragmented and thus limiting the number of cells in each subcluster, which reduces its power. Do you have significantly more "normal_like" cells than "epithelial_C"?

As for whether it is normal of not to not have chromatic alteration in your results, that is a totally possible outcome. Not all cancers involve large chromosomal rearrangements, and even within a certain type of cancer, it can vary by specific subtype.

Regards, Christophe.