Assembly of heterozygous genomes

Heterozygous genome assembly remains a complex task for which no satisfying solution exists. Phasing chromosomes in diploid or polyploid species or in metagenomes is still an open problem, despite the emergence of new... [ view full abstract ]

Heterozygous genome assembly remains a complex task for which no
satisfying solution exists. Phasing chromosomes in diploid or
polyploid species or in metagenomes is still an open problem,
despite the emergence of new long-read technologies and other
dedicated approaches. In particular, intra-species or
inter-species variations are usually discarded and/or result in
highly fragmented assemblies.

In this work we propose Bwise, a novel de Bruijn graph-based
assembler that fully takes advantage of paired-end relationships
between reads, thereby allowing correct paths to be determined
while resolving haplotypes and genomic repeats. Preliminary
results on simulated datasets show that Bwise prevails the
Spades state-of-the-art assembler in particular on
hard problems: when heterozygosity rate is medium (~1%). Results
also show that Bwise uses limited ressources (a human genome may
be assembled using 50GB memory).

Beyond paired-end reads, the proposed framework allows in
principle the integration of long-range information (mate pairs,
3Cseq/Hi-C, Single Molecule Real-Time (SMRT) as well as Nanopore
reads) to determine accurate long paths in the assembly graph,
with the ultimate goal of generating a single high-quality contig
per chromosome.