Abstract

Encoding the genetic diversity of a species into a ³graph reference², with applications to antibiotic-resistance and antigenic variation

The choice of reference object used to represent the genome of a species has far reaching effects on biological inference. The use of a ³reference genome² made pragmatic sense when sequencing was expensive. However it can also lead to under-detection of non-reference alleles and various artefacts. The problem is substantial in regions of genomes of high diversity, such as the human MHC , or where unusual mutational mechanisms occur, as observed in the var genes in P. falciparum, for instance. Indeed, the limitations of a single reference genome have long been recognized in bacterial genomics, where the term pan-genome has been coined to refer to the full set of genes found across all individuals of a species. Recently progress has been made towards using an ensemble of genomes or variation data as a first-class object . However many algorithmic challenges remain and few implementations exist. I will discuss two lines of research from my group, based on the idea of building a reference graph of known sequence and variation in a species. The first relates to the prediction of antibiotic resistance in bacteria, recently published for S. aureus and M. tuberculosis (nature.com/ncomms/2015/151221/ncomms10063/full/ncomms10063.html). The second shows how a new data structure based on a generalised Burrows-Wheeler Transform and FM-index allow encoding of a graph of variation and read-mapping to the graph. I will show the impact of this method on the study of antigenic diversity in P. falciparum, the parasite responsible for the most deadly form of malaria.