Detecting copy number alterations based on RNA-seq data

Question

[igor]

I've seen a few papers use RNA-seq data to detect CNVs. However, it's always using "custom scripts" in my experience. Is there a publicly available tool for performing that analysis?

I found inferCNV, but that is designed for single-cell data which will have a very different profile from bulk RNA-seq and have a lot more replicates.

Answer 1

[igor]

I just found HoneyBADGER, another method for single-cell RNA-seq:

HoneyBADGER (hidden Markov model integrated Bayesian approach for detecting CNV and LOH events from single-cell RNA-seq data) identifies and infers the presence of CNV and LOH events in single cells and reconstructs subclonal architecture using allele and expression information from single-cell RNA-sequencing data.

[igor]

There is also CONICS:

CONICSmat is an R package that can be used to identify CNVs in single cell RNA-seq data from a gene expression table, without the need of an explicit normal control dataset. CONICSmat works with either full transcript (e.g. Fluidigm C1) or 5'/3' tagged (e.g. 10X Genomics) data.

[igor]

Yet another option is CaSpER:

CaSpER is an algorithm for identification, visualization and integrative analysis of CNV events in multiscale resolution using single-cell or bulk RNA sequencing data.

[igor]

Another option: https://github.com/jmonlong/scCNAutils

Answer 2

[bhaas]

The InferCNV software aims to make the Tirosh & Regev method available as a toolkit:

https://github.com/broadinstitute/inferCNV/wiki

Answer 3

[Chris Miller]

To my amazement, Aviv Regev's lab did it, but I cannot find the paper at the moment (please link it here if you find it!). Suffice to say, it's a complex process, and one which produces only very rough outlines of the CN landscape. You will probably only pick up very large events (multi-Mb), and it will likely depend on having a large cohort with which to average out the noise inherent in expression.

[cbst]

Is this the paper you are referring to?

[Chris Miller]

Yeah, I saw Aviv talk at a conference, but that seems like the right paper. Even more flabbergasted that it works with single-cell RNAseq. Amazing stuff!

Answer 4

[Thind amarinder]

We summarized a list of such tools in our review published in BIB. Now there are 4-5 tools to call CNA from bulk RNAseq.

[igor]

The relevant section: "recent advancements in bioinformatic tools and methods have assisted in the determination of CNA from bulk and single-cell RNA-Seq (scRNA-seq) data. Some of bulk RNA-Seq-based tools are CaSpER [29], CNAPE [30], CNVkit-RNA [31] and SuperFreq [32]."

Answer 5

[igor]

There is now an RNA-seq version of CNVkit for bulk RNA-seq data.

For a subset of genes, mostly housekeeping genes, the mRNA expression levels measured by transcriptome sequencing are mostly explained by underlying the genic regions’ genomic copy number. CNVkit can use this information to estimate coarse-grained copy number from RNA sequencing of a process-matched cohort of samples.