mcveanlab / corticall

Germline DNM caller based on LdBG assembly of pathogens

Corticall: A graph-based de novo mutation caller based on traversing (Mc)Cortex de novo assembly graph and link data.

Quick start

git clone https://github.com/mcveanlab/Corticall
cd Corticall
ant
java -jar dist/corticall.jar -h

Introduction

Corticall is a Java-based de novo mutation (DNM) caller based on linked multi-color de Bruijn graphs (LdBG) produced by McCortex. It is able to leverage multiple reference sequences to better characterize DNMs in otherwise inaccessible regions of canonical reference genomes.

Corticall may also be used as a class library for performing efficient, low-memory traversals on LdBGs. The most important functionalities provided are:

• iterating over records in a Cortex graphs
• random access (by binary search) to Cortex graph records
• performing simple walks (i.e. extracting a contig, optionally using links to disambiguate junction choices)
• performing walks assisted by one or more reference sequences in a manner consistent with link information
• performing depth-first searches with custom stopping rules (useful for finding interesting graph motifs)
• aligning k-mers and contigs back to reference sequences

Corticall handles the heavy lifting when operating with these data structures, permitting developers to concentrate on the genome analysis and variant calling tools that can be written on top of this API.

Availability

Corticall is released under the Apache 2.0 license. The latest code is freely available at Github.

Dependencies

Corticall has the following dependencies:

• Java8: needed for runtime and development kit
• Apache Ant: for dependency fetching and compilation

Additional recommended software:

• McCortex: for building Cortex graphs and link annotations

Installation

To download and compile:

git clone https://github.com/mcveanlab/Corticall
cd Corticall
ant

To run (and get a listing of available commands):

java -jar dist/corticall.jar

To get help for a specific command (e.g. "Call"):

java -jar dist/corticall.jar Call

Support

Please contact Kiran V Garimella ([email protected]) with any questions/comments/concerns/cake. Feedback, bug reports, and pull requests are welcome.

Related projects

For a pythonic take on programmatically accessing McCortex linked multi-color de Bruijn graphs, see Winni Kretzschmar's cortexpy project.

Citations

The original multi-color de Bruijn graph paper can be found at:

• "De novo assembly and genotyping of variants using colored de Bruijn graphs", Iqbal, Caccamo, Turner, Flicek, McVean (Nature Genetics) (2012) (doi:10.1038/ng.1028) http://www.ncbi.nlm.nih.gov/pmc/articles/PMC3272472

Our manuscript on linked multi-color de Bruijn graphs is now available via Bioinformatics at:

• "Integrating long-range connectivity information into de Bruijn graphs", Turner, Garimella, Iqbal, McVean (Bioinformatics) (2018) (doi:10.1093/bioinformatics/bty157) https://www.ncbi.nlm.nih.gov/pubmed/29554215

A manuscript preprint for Corticall, our de novo assembly approach to de novo mutation detection in pathogens, can be found at:

• "Detection of simple and complex de novo mutations without, with, or with multiple reference sequences", Garimella, Iqbal, Krause, Campino, Kekre, Drury, Kwiatkowski, Sa, Wellems, McVean (2019) (doi:10.1101/698910) https://www.biorxiv.org/content/10.1101/698910v1

corticall's Issues

Does CortexJDK offer a low-memory way to merge sorted cortex graphs?

If so, is there a command line hook for that?

ctxgraph usage: short arguments should be short

Please rename -ieg command line argument to -g. Also update usage from

Graph with inferred edges

to

Sorted cortex graph with inferred edges

and

Contigs (in FASTA format)

to

Contigs (in FASTA format - must be indexed with samtools)

where can I find cortexjdk

Hi,
I am searching for cortexjdk to find. I tried to compile the master branch with ant. it produces corticall.jar. I cant find tools like RecoverExludedKmers or Join.

Thanks,
Moustafa

ctx spec

I think "The first uint64_t of a kmer_container is always partially empty unless the stored kmer fits exactly into the kmer_container." may be incorrect.

I think to be correct, the second word needs to be changed from first to last.