Many human traits are strongly influenced by our DNA. E.g., genetic variants determine whether we tend to be taller or shorter than average, or whether we are more or less likely to develop diseases. A Genome-Wide Association Study (GWAS) is a statistical tool for finding which genetic variants are most important for a particular trait. Over the last decade, GWAS have identified DNA variants for thousands of human traits.

While successful, existing tools for performing GWAS tend to be computationally demanding. For example, one of the most popular tools, BOLT-LMM, can take weeks to analyze data for 400,000 individuals. To overcome this problem, Jasper Hof and Doug Speed developed LDAK-KVIK, a new GWAS tool that is substantially more efficient than existing tools. Specifically, LDAK-KVIK has three main benefits. Firstly, it only needs to read in data for 250 genetic variants at a time, meaning that its computational demands are much lower than existing methods, which read in up to one million genetic variants. Secondly, LDAK-KVIK incorporates advanced models for the expected importance of different genetic variants, whereas existing tools assume all variants contribute equally. Thirdly, LDAK-KVIK can both test individual genetic variants and sets of genetic variants, and therefore can more precisely locate the regions of DNA influencing different traits.

LDAK-KVIK has just been published in the journal Nature Genetics. The paper starts by explaining the computational tricks used by LDAK-KVIK to reduce both runtime and memory usage. Then it compares LDAK-KVIK with four existing GWAS tools. The results show that LDAK-KVIK can analyze data for 400,000 individuals in less than ten CPU hours, meaning it is orders of magnitude faster than the closest rival, BOLT-LMM. Finally, the paper analyzes 40 complex traits recorded within the UK Biobank, and demonstrates that LDAK-KVIK consistently finds more significant genetic variants than the four existing GWAS tools (see figure above).

The first author, postdoc Jasper Hof from QGG, explains: "By developing LDAK-KVIK, we have set a new benchmark for statistical software that analyze GWAS data. When developing this method, Doug and I realized how difficult it is to create a software that can robustly analyze large GWAS datasets. I am proud that we achieved this, and I am excited about the prospect of using our developed software to tackle new challenges in statistical genetics".

In addition to the publication, the authors have also created the website www.ldak-kvik.com, which contains details of how to download and run the new tool.