Skip to main content
Premium Trial:

Request an Annual Quote

Decode Genetics Preprint Suggests Olink Outperforms SomaLogic, But Experts Point to Study Flaws


NEW YORK – A recent study from Decode Genetics has sparked controversy in the world of plasma proteomics by suggesting that Olink's Explore platform outperforms SomaLogic's SomaScan system.

Scientists not involved with the work, as well as a SomaLogic official, have disputed the study's findings, claiming flaws in its experimental design and procedures and calling into question whether the data presented in fact demonstrates the superiority of Olink's system.

Uppsala, Sweden-based Olink and Boulder, Colorado-based SomaLogic are the two dominant providers of affinity-based assays for high-throughput, highly multiplexed plasma proteomics. Olink's Explore platform measures up to 3,000 proteins via the company's proximity extension assay technology, which uses pairs of antibodies linked to DNA strands that are brought into proximity when the antibodies bind and are then extended by a DNA polymerase, creating a new sequence that can be used as a surrogate marker for the target protein. SomaLogic's SomaScan measures roughly 7,000 proteins using the company's Somamer reagents, a form of aptamer.

Few studies have directly compared the two platforms. Perhaps the most notable is a project published last year in Nature Communications by a University of Cambridge-led team that included SomaLogic Chief Medical Officer Stephen Williams. That study did not seek to assess which, if either, of the two platforms was superior but did see substantial divergence between their data. Specifically, it found that roughly 36 percent of the protein-phenotype links were observed by only one of the two systems, indicating, the authors wrote, "that both techniques capture distinct aspects of protein biology."

The Decode study, which was published last month as a BioRxiv preprint, looked at results from a pair of large-scale proteogenomic projects performed using either of the two platforms. In one, researchers at the UK Biobank ran samples from 47,151 European subjects on Olink's Explore 1536 platform (which measures 1,459 proteins). In the other, Decode researchers ran samples from 35,559 Icelandic subjects on the 5,000-protein version of the SomaScan platform.

In both studies, the researchers looked for protein quantitative trait loci, or pQTLs, meaning links between genetic variants and plasma protein levels. PQTLs have emerged as a major focus of research done on the Olink and SomaLogic platforms as such linkages can help researchers, for instance, identify promising drug targets or patterns of protein expression linked to the development of disease. The Olink and SomaLogic platforms have enabled such research by providing the proteomic coverage and throughput needed to find such linkages in large-scale population studies.

PQTLs are typically characterized as either cis — meaning that the pQTL is located close by the gene that encodes that protein — or trans, meaning it is located further away from the gene encoding the protein. While both are potentially meaningful, detection of a cis pQTL provides strong indication that the platform's assay for that particular protein is specific — that it is, in fact, measuring the target it is meant to.

Specificity is an inherent concern for highly multiplexed affinity assays like those offered by Olink and SomaLogic. In their preprint, the Decode researchers found that a large majority (84 percent) of the assays on the Olink Explore platform has cis pQTLs, indicating that they were detecting their stated targets. The researchers found that only 38 percent of the SomaScan assays had cis pQTLs.

The findings play into pre-existing questions about the specificity of the SomaScan platform. Olink's Explore — as well as most other antibody-based platforms — uses two antibodies to detect target proteins, which improves assay specificity. SomaLogic's Somamer-based approach, on the other hand, uses only a single capture agent for each protein. This has led to concerns in some circles about the technology's specificity and has also become a talking point for Olink in discussions of the two platforms and their capabilities. For instance, in an interview last month prior to the release of the Decode study, Olink President and CEO Jon Heimer highlighted "specificity" as "the main important point where we are quite different."

Somamers are a modified form of aptamers, nucleic acid-based reagents capable of binding proteins or other targets. These can then be read out via technologies like PCR, microarrays, or next-generation sequencing. SomaLogic has worked to combine two forms of specificity into the reagents by developing the molecules so that they would exhibit slow dissociation rates from their targets — on the order of an hour or more — in the case of true hits and fast dissociation — on the order of seconds — in the case of off-target binding.

This, combined with the molecules' affinity for their targets, provides two dimensions of specificity in a single molecule, which the company has said means they can offer highly specific detection without requiring a secondary detection molecule as in a sandwich immunoassay.

SomaLogic CEO Roy Smythe disputed the Decode researchers' results and the concerns about the SomaScan system's specificity more generally. He said that over the least 18 months, the company has been going through its assay menu and doing mass spectrometry pull-down experiments to determine that its Somamers are binding to their intended targets. So far, it has made it through roughly half its menu, Smythe said, and noted that for roughly a third of the targets for which the Decode study detected no cis pQTL, SomaLogic has pull-down data showing its assays correctly hit those proteins. He also said that SomaLogic is exploring the use of two Somamers for some protein targets to address concerns about specificity.

Smythe added that the study compared results from the two platforms used in two different populations, noting that, ideally, a comparison would look at the results in the same study cohorts.

"You are taking two different sample sets from two genetically different populations," he said. "We also actually know that there are some issues with sample handling that could impact proteomics for one of these two sample sets. So you are running two different proteomic measurement approaches on two different sample sets with characteristics that are neither controlled for nor overtly discussed [in the paper] as a potential drawback for the study."

He also said that the Decode researchers had not normalized the SomaScan data in the manner SomaLogic recommends, noting that failure to do so could lead to the omission of biological events including cis pQTLs.

Claudia Langenberg, a researcher at the University of Cambridge and the Berlin Institute of Health at Charité, and the senior author on the 2021 Nature Communications paper looking at the two platforms, said she thought the Decode preprint, which has not undergone peer review yet, was "a pretty bad paper."

"I'm quite surprised because [Decode CEO and study senior author] Kari Stefansson is an amazing scientist," she said. "That he let that [paper] go out is quite surprising."

Langenberg and Smythe said they had heard the Decode group was considering taking down the preprint. A research note from investment bank Stifel also suggested the paper could be removed from the preprint server. According to BioRxiv policy, manuscripts posted on the site cannot be removed, though authors can mark their papers as "withdrawn" if they believe the work is flawed. Stefansson did not respond to requests for comment on either the study or the possibility of Decode having it marked "withdrawn."

Langenberg, who works with both systems, noted that when looking only at the proteins measured by both platforms, the gap in cis pQTLs narrows, with 87 percent of Olink assays detecting a cis pQTL versus 59 percent for SomaLogic assays. She suggested that this comparison was the more relevant one and should have been highlighted more prominently in the paper. Another consideration, she suggested, was the fact that while a higher proportion of Olink assays detected cis pQTLs, the SomaScan platform, with its larger assay library, still detected more cis pQTLs overall — 84 percent of 1,459 for Olink versus 38 percent of 4,907 for SomaLogic. SomaLogic's platform currently has assays to 7,000 proteins versus 3,000 for Olink. The two companies plan in the near future to expand their platforms to 10,000 and 4,500 proteins, respectively.

"If you start with 7,000 and you get 50 percent [pQTL hits], that is more than if you got 50 percent from 3,000," Langenberg said. "So that is a consideration. And [the Decode researchers] didn't really talk about that. They just talked about the relative number."

Langenberg agreed with Smythe that normalization of the SomaLogic data according to the company's recommendations improves cis pQTL detection. She cited comparisons her teams have done of their SomaScan data, which is normalized, with data from Decode, which is not normalized.

"Even though the Decode [dataset] is almost three times as big, we have cis pQTLs using the normalized data that they don't see," she said.

She said, however, that the need to normalize the SomaLogic data "is a headache," noting that it makes it difficult to compare results across platforms and, unless all labs are normalizing their data to compare results from SomaScan experiments that were performed by different groups.

"If you have an analysis where normalized data versus non-normalized data gives a very discordant set of results, that's not great," Langenberg said. "It makes cross-study comparison very difficult."

She said that her group has decided it will only report associations it sees in both normalized and non-normalized SomaScan data, "because otherwise how can you validate, how can you replicate?"

Beyond the scientific question, there is also the consideration of what, if any, impact the Decode paper could have on the two companies' stock prices. Both Olink and SomaLogic went public last year. Research notes from investment banks have been mixed, with SVB Leerink analyst Puneet Souda characterizing the paper as "pushing Olink ahead in the high-plex proteomics race," while Stifel analyst Daniel Arias was more skeptical of the Decode results, citing outside researchers who raised concerns similar to those of Smythe and Langenberg.

SomaLogic's shares fell more than 10 percent in the days following the release of the preprint on Feb. 20. In Tuesday morning trading on the Nasdaq, the shares were at $8.26, down 12 percent from $9.40, its closing price on the last day of trading before the paper was posted. Over that same period, Olink stock was up 9 percent from $15.74 to $17.09.