Zamin Iqbal, who has been doing fascinating work with development of Cortex genome assembler, sent us an update with link to his latest paper published in Bioinformatics. Here is the abstract.
We have developed a software package, Cortex, designed for the analysis of genetic variation by de novo assembly of multiple samples. This allows direct comparison of samples without using a reference genome as intermediate, and incorporates discovery and genotyping of SNPs, indels and larger events in a single framework. We introduce pipelines which simplify the analysis of microbial samples and increase discovery power; these also enable the construction of a graph of known sequence and variation in a species, against which new samples can be compared rapidly. We demonstrate the ease-of-use and power by reproducing the results of studies using both long and short reads.
Cortex is an unusual genome assembler. Almost all other assemblers (no matter whether de Bruijn graph-based, or non de Bruijn graph-based) start with the assumption that two pairs of diploid chromosomes are identical. That is a pretty good assumption except where it does not work :). Recently there is lot of interest in finding information about phasing and genomic variation among a large population. Aligning reads back to reference genome is the most used strategy to identify SNPs, but insertion-deletions are harder to find with typical alignment methods. Alignment strategy has a strong bias toward finding what one already knows, i.e. the reference genome.
Based on comments from others in Twitterosphere and elsewhere, Cortex is very good program not only for the originality of its algorithm (‘colored’ de Bruijn graph), but also because it is light-weight and scalable. We expect to go through the code and algorithm in the coming months, but for the time being, we have to leave you with Zamin’s email comment.
We have a new paper showing we can distinguish 2 strains of MRSA in 3 minutes and 160Mb of RAM, and get as good results as a PNAS publication that did whole genome assemblies and alignment. That was with Sanger reads, which is kind of surprising that with long reads a de Bruijn approach can do comparably well as a finished WG consensus assembly + alignment. Then we take 72 strains of S. aureus (100x 100bp illumina reads for each one) from another paper, and again reproduce their results (10Gb RAM, 5 hours serially, I forget how long if in parallel).
The goal was
1. to show that you can go from reads to variants very quickly now.
2. it provides a new way to compare a new sample with all known previous samples, by comparing with the graph of all previous samples