The advancement in sequencing technologies is producing an acceleration in the acquisition of genomic information never seen before. Population-scale human genome sequencing initiatives are ongoing in several nations with the main objective of producing a comprehensive catalogue of genetic variants within a population, for developing personalized medicine at national level. Computer scientists have contributed by laying the algorithmic foundations of computational graph pangenomics, with the goal of effectively leveraging the unprecedented opportunity of profiling genomes at population scale. This contribution is only in its initial stage, since the routine assembly of pangenomes and the construction of graphs representing them is showing limitations of current state-of-the-art tools for managing and analyzing large-scale pangenomic data. Limitations are not only due to the huge amount of data, but also to genomic differences in the scale to which mutations occur in different organisms (e.g. eukaryotes versus microbial), and in the size (small and large variations, SNPs versus SVs) of mutations acquired. Since pangenomes are inherently more complex than single genomes, analyzing them is still hard.

This project builds upon the idea of representing pangenomes as labeled graphs to develop a new computational framework, Multiscale Pangenomics: the novel idea is moving away from a single graph “dogma” towards an integrated collection of graph representations, based on different genomic scales, and on multiscale algorithms, building and considering a graph with the most appropriate resolution. This will allow to couple high accuracy with quick turnaround time, limiting the computational resources needed and making it feasible to attack some of the most challenging problems in bioinformatics such that the reliable identification of mutations causing rare diseases, where massive pangenome graphs help by removing redundancies without discarding individual differences.