Question: Which Driver-Gene Detecting Software Supports hg38?
gravatar for sviatoslav.kendall
9 weeks ago by
United States
sviatoslav.kendall510 wrote:

I'm interested in detecting driver genes in some cancer biopsies that have been sequenced using hg38 as the reference. I've looked at three programs so far (MutSig2CV, MuSiC and 2020+) and found clear indications that two of these only support hg19 and I have not found an answer in the online documentation for the third (2020+).

Although I am most interested in knowing which software currently supports hg38, I wouldn't mind hearing about planned upgrades and work-arounds.

music 2020+ mutsig2cv • 212 views
ADD COMMENTlink modified 4 weeks ago by arvind.shankar310 • written 9 weeks ago by sviatoslav.kendall510

I'm interested in detecting driver genes in some cancer biopsies

...but has this not been done already by other groups? Is it for a Masters project or some other form of training?

I would directly contact the authors of these programs in order to find out whether or not they plan to support hg38 / GRCh38 or not.

ADD REPLYlink written 9 weeks ago by Kevin Blighe46k

I'm going to be looking at cancers that haven't been studied so thoroughly.
I've already reached out to either the authors or people involved with maintaining the software I mentioned to ask about work-arounds and planned upgrades - haven't heard back from anyone yet. I thought this issue might have some mass appeal because the programs I mentioned appear to be some of the most popular for doing this type of work and I was surprised to learn they don't support the latest build of the human genome even though it's already more than five years old.

ADD REPLYlink written 9 weeks ago by sviatoslav.kendall510

It is a recurring theme in bioinformatics for developers to go 'quiet' and the software to go un-maintained - the reason could be lack of funding and/or the person moved job. Do you actually need these programs? Why not identify driver mutations and genes through a manual data processing pipeline?

ADD REPLYlink written 9 weeks ago by Kevin Blighe46k

I'm the person that developed 20/20+ and I have moved to another lab. I do still maintain the software. But all improvements are done in my spare time without really any recognition for doing so. I would, however, recommend against "manual data processing pipelines" as such ad hoc approaches are why the cancer literature is filled with contradictory (and often wrong) information.

ADD REPLYlink modified 7 weeks ago • written 7 weeks ago by Collin680

Hey Collin - no worries. I will also go quiet some day when I have moved on from here!

ADD REPLYlink written 7 weeks ago by Kevin Blighe46k

I've been a secret fan of your responsiveness to the BioStars community for a while. I think it's a great service to the bioinformatics community

ADD REPLYlink written 7 weeks ago by Collin680

Regarding your initial question, 20/20+ currently (as of 6/29/2019) does not support hg38, but we hope to implement it soon.

ADD REPLYlink written 7 weeks ago by Collin680
gravatar for Inigo Martincorena
8 weeks ago by
Sanger Institute, Cambridge, UK
Inigo Martincorena190 wrote:

Hello. dNdScv, which I developed, supports GRCh38. Although it uses hg19 by default, the package contains the files and instructions to run GRCh38, as well as functions to run it on any assembly or species. In benchmarking analyses by PCAWG (the ICGC/TCGA pancancer analysis of whole genomes consortium) dNdScv was shown to perform very similarly to MutSigCV in terms of specificity and sensitivity.

dNdScv is an R package. To install it you have to use:

library(devtools); install_github("im3sanger/dndscv")

To run driver discovery on GRCh38, you need to download the precomputed database for GRCh38, and then use:

dndsout = dndscv(mutations, refdb="RefCDS_human_GRCh38.p12.rda", cv=NULL)

See the tutorials above for more details if you are interested in the method.

ADD COMMENTlink written 8 weeks ago by Inigo Martincorena190

Hi, I'm Collin, one of the first authors on the TCGA PancanAtlas drivers paper. When we benchmarked several methods, MutSigCV did ok, but had several cancer types where the results were flagged as substantial outliers compared to all the other methods. We however did not benchmark your method.

ADD REPLYlink written 7 weeks ago by Collin680
gravatar for arvind.shankar3
4 weeks ago by
arvind.shankar310 wrote:


I am working on something similar, and know 3 farily recent tools that work on hg38.


  2. (The one Inigo suggested)

  3. The third one doesn't really need any specific genome build.

Note: All of these infer selection in cancer genomes, you might have to do a little bit more downstream analysis for cancer drivers. In general all positively selected genes are potential cancer drivers.

ADD COMMENTlink modified 4 weeks ago • written 4 weeks ago by arvind.shankar310
Please log in to add an answer.


Use of this site constitutes acceptance of our User Agreement and Privacy Policy.
Powered by Biostar version 2.3.0
Traffic: 1810 users visited in the last hour