Alignment of next-gen sequence data and variant discovery for large Dutch cohorts

Processing sequence data of the Dutch genome project (GvNL) for (a) the final step of read alignment; (b) calling genetic variants at the population level. The Dutch Genome Project (Genoom van Nederland, GvNL) aims to build a deep catalog of all types of genetic variants within the Dutch population. In this project the whole genome of 250 trios (750 individuals) have been sequenced. The data generated consists of billions of short sequence reads, which need to be processed per participant and the evidence of variant will be summarized at population level. The reads in each individual were first aligned against a reference genome while the major part of this computationally intensive work is being performed on local clusters and on the life science grid. The sequence reads per individual are stored as several pairs of files and the alignment results of all pairs must be merged to summarize all the information. Unfortunately, the last step of this process, merging of multiple alignment files of one person, cannot be done on those resources of local clusters and on the life science grid due to lack of temp diskspace. The second part of this project (population-scale variant calling of SNPs, indels and structural variants) requires both storage and processing power as well as fast connection between them. We foresee that Huygens would be an ideal for this type of applications (e.g. Pindel).