Frédéric Delsuc - Genomics from roadkill - high quality mammalian genomes using hybrid assembly with MinION long reads
London Calling 2019
With thousands of fatalities due to car collisions with wildlife reported each year, roadkill are an underexploited resource in genomics. Here we show that mammalian roadkill samples could be used as a suitable source of DNA for long-read sequencing using the MiniON device for two carnivoran species frequently encountered along South African roads: the bat-eared fox (Otocyon megalotis) and the aardwolf (Proteles cristatus). For both species, hybrid assembly of 150PE Illumina reads at ~85X coverage (~215 Gb) and MiniON long reads at ~12X coverage (~30 Gb) using the MaSuRCA assembler provided genomes with high contiguity (~10,000 contigs with N50 of ~700 Kb) and completeness (>90% of complete BUSCOs). We further demonstrate that about 90% of the 14,509 single-copy orthologous genes of the OrthoMaM database could be successfully retrieved from these assemblies. These figures compare favourably with current mammalian genome assemblies and set our genomes among the best carnivore genomes currently available. This cost-effective strategy to obtain high quality reference mammalian genomes opens the way for large-scale population genomic studies of mammalian wildlife using resequencing of samples collected from roadkill. We illustrate the potential of the approach for genome scale species delimitation in both species for which subspecies have been defined based on disjunct distributions and morphological differences.
In the final talk of the session, Frédéric Delsuc from the French National Centre for Scientific Research began by talking about convergent evolution, the process by which similar phenotypes arise independently in phylogenetically distant lineages. The well studied examples being the case of placental and marsupial mammals, where certain features have evolved to fill the same environmental niches independently. Examples of convergent evolution include: the mole and the marsupial mole; the mouse and the marsupial mouse; the wolf (placental) and the Tasmanian wolf (marsupial). Frédéric went on to introduce the ConvergeAnt project studying convergent evolution of ant eating placentals where this similar morphology of ant eaters has 5 different independant [sic] origins. Frédéric plans to sequence 5 reference genomes using Nanopore technology to study how genome sequence reflects this convergent evolution. However, many of these species are rare, elusive and protected, and many are unfortunately frequently involved in road traffic accidents - yet this roadkill could serve as a source of tissue for these projects.
Frédéric started the second part of his talk discussing his work sequencing the genomes of roadkill animals with the MinION. Despite the challenges of working with post mortem samples (unknown sample age and DNA degradation), his team were able to sequence the genomes of two carnivoran genomes, the Bat-eared fox (Otocyon megalotis) and Aardwolf (Proteles cristatus), on the MinION, generating 33 Gb (13x coverage) and 29.4 Gb (11x coverage) of data, respectively. Frédéric shared some tips for isolating long fragments from ear samples from roadkill: roadkill samples stored in RNAlater rather than 95% ethanol reduced degradation, and removing necrotic cells with a scalpel under a microscope also helped isolate more intact DNA fragments.
The nanopore reads were combined with short read (~80x coverage) data to build a hybrid assembly. Assembling using MaSuRCA produced assemblies with >40-fold fewer contigs and a >30-fold increase in contig N50 compared to a previous short read only assembly (SOAPdeNovo). The Bat-eared fox assembly had 12,735 contigs with an N50 of 676 kbp. The Aardwolf assembly had 8,874 contigs with an N50 of 699 kbp. Compared to the existing assemblies, these assemblies rank among the best carnivore genomes currently available for contiguity and completeness.
Frédéric concluded his talk by stating that roadkill is a suitable source of DNA for Oxford Nanopore sequencing, and that combining 10X coverage of MinION nanopore reads with short reads and is a cost-effective strategy to obtain high quality reference mammalian genomes with high contiguity and completeness for less than $15,000.