NEW YORK — DeCode Genetics CEO Kári Stefánsson kicked off the annual European Society of Human Genetics meeting this week with an overview of how the Icelandic company is blending proteomics and genomics to advance its internal disease research programs.
The meeting, originally scheduled to take place in Vienna, was moved to a virtual platform because of the ongoing COVID-19 pandemic.
Reykjavik-based DeCode has been using SomaLogic's SomaScan platform to measure the levels of nearly 5,000 proteins in individuals from its biorepository, Stefánsson said, and has been correlating those measurements with data on sequence variants linked to diseases and other traits. So far, the company has identified markers associated with osteoarthritis and autoimmune thyroid disease and is developing a calculator to predict five-year mortality risk.
According to Stefánsson, two decades of interrogation of the human genome using tools like microarrays and next-generation sequencing has resulted in the discovery of a trove of sequence variants associated with diverse clinical phenotypes. Yet linking a variant to a phenotype remains challenging and studying the plasma proteome might help to bridge that gap.
"The proteome has an advantage over the transcriptome in that basically all proteins in the body make it into blood," said Stefánsson. He noted that the study of plasma proteins can enable researchers to find the causal gene behind a phenotype and underscored that as protein levels rise and fall over time, they offer a temporal glimpse into disease regression and progression.
Scientists who interrogate the proteome need to do so in a hypothesis-independent manner, though, he stressed, an approach made possible by the availability of protein-profiling platforms such as those offered by Boulder, Colorado-based SomaLogic, or Uppsala, Sweden's Olink.
Stefánsson noted that Olink currently enables the measurement of about 1,500 proteins using its proximity extension assay technology, which uses Illumina sequencing for readout. SomaLogic's assay, in contrast, relies on protein-capture reagents called SomaMers to measure proteins, which are then quantified using hybridization microarrays.
Stefánsson pointed out that most of the proteins available on SomaScan were not available on Olink at the time of DeCode's study, though he added that both companies have recently upgraded their offerings, with SomaLogic now capable of measuring 7,000 proteins, while Olink will be able to measure about 3,000 proteins in coming weeks.
DeCode had previously identified the association of roughly 265,000 genetic variants and about 4,700 proteins with 373 diseases and other traits, and about 80 percent of those proteins were not included on Olink, Stefánsson said, so the firm opted to use SomaScan. He noted that most of the work DeCode did on the SomaScan platform used an earlier generation assay that measured 5,000 proteins, and the company previously announced agreements with SomaLogic in 2018 and in 2020.
"The more proteins you look at, the less you need to rely on a priori assumptions," he said.
As part of its study, DeCode obtained SomaScan protein measurements in roughly 36,000 Icelanders and then compared the levels of those proteins with 373 diseases and other phenotypes. They also matched them to about 27 million sequence variants. The cohort DeCode used was collected over the past 20 years, and while about 80 percent of the Icelanders were still living, 20 percent have since passed away.
Some new findings were in osteoarthritis, autoimmune thyroid disease, and mortality prediction.
For osteoarthritis, Stefánsson noted that there is a "desperate need for a good biomarker for onset of disease and progression." DeCode's scientists discovered that the level of the protein-coding gene CRTAC1 is strongly associated with many types of the disease, he said.
Moreover, DeCode found that the higher the concentration of the protein, the more likely a subject was to undergo joint replacement within a given timeframe. Stefánsson believes the marker could serve as an indicator of disease progression. "That could be of substantial use when it comes to clinical development of potential therapy for the disease," he said.
He also discussed a meta-analysis of autoimmune thyroid disease in Icelanders and British subjects in the 500,000-sample UK Biobank. The analysis covered roughly 30,000 cases and 725,000 controls and resulted in the identification of 99 sequence variants and 93 loci, of which 84 were novel. The intronic variant FLT3 was the "strongest hit," he said, and further research showed association of the variant FLT3 with rheumatoid arthritis, celiac disease, and systemic lupus erythematosus. The variant also increases risk of acute myeloid leukemia, he added.
As roughly 20 percent of the cohort surveilled in the study had died since recruitment, DeCode's scientists also looked at the protein measurement data to see if a combination of proteins might enable them to predict when someone might die.
By pairing data on about 100 proteins with other risk factor information, DeCode was able to predict that 5 percent of the people aged 60 to 80 had about a 90 percent probability of dying within five years, and that 5 percent had a very low probability of dying.
"This is a powerful calculator to predict five-year mortality risk," Stefánsson said.
While admitting such a mortality calculator "looks Orwellian" on its surface, he said the firm had already applied the calculator to a clinical trial, where it could show those on an active drug had a longer predicted lifespan, versus those who had received a placebo. "This might actually turn out to be useful as a test that could be used in all kinds of clinical development," said Stefánsson.
In a follow-up email, he said that DeCode had not yet published these findings but that a paper was currently being reviewed for publication. Stefánsson cofounded DeCode Genetics in 1996, and Amgen acquired the company in 2012 for $415 million.