Skip to main content
Premium Trial:

Request an Annual Quote

Sage Bionetworks Challenging Teams to Develop Predictor for Anti-TNF Therapy Response in Rheumatoid Arthritis


Starting this fall, researchers will use an online system developed by Sage Bionetworks to at once collaboratively share data and compete to develop a genetic predictor that can identify rheumatoid arthritis patients who are likely to respond to immunosuppressive anti-TNF therapies.

The effort, called the Rheumatoid Arthritis Responder Challenge, is one of several challenges opened by Sage in collaboration with the Dialogue on Reverse Engineering Assessment and Methods, or DREAM, project this year that will rely on its Synapse cloud-based data analysis platform. Last week, the RA challenge was highlighted in a White House Office of Science and Technology Policy event honoring Sage President Stephen Friend as one of 13 "Open Science Champions of Change."

In the context of the RA challenge, Sage and the researchers involved are hoping that using Synapse to share and combine datasets that would otherwise be inaccessible or disparate will yield a result that has previously been elusive.

Currently, up to a third of rheumatoid arthritis patients do not respond to the standard course of anti-TNF treatment and there are no existing genetic predictors to identify who will or will not be in that non-responsive group, Robert Plenge, director of genetics and genomics in the division of rheumatology, immunology and allergy at Brigham and Women's Hospital, told PGx Reporter this week.

Plenge and his colleagues have been doing research into anti-TNF therapy response for several years, most recently completing a GWAS study of over 2000 RA patients, which the team published earlier this year in PLoS Genetics.

"With the GWAS data a couple of things happened," said Plenge. "First, we weren’t identifying any single signal of association in a [cohort] we'd consider large for a pharmacogenetic study. Second, we thought there might be a polygenic signal after doing some additional modeling, but we couldn't distinguish the signal from the noise."

Plenge said both his group and Sage shared the idea of launching a challenge where research teams could attempt to extract a signal by integrating the GWAS dataset with other available datasets, like gene expression, protein-protein interaction, HapMap cell line data, and some RNA-seq data. “People could try different approaches … to try to solve this question," he proposed.

According to the challenge team, funding and support to carry out the effort and to support Sage's Synapse system is coming from the Arthritis Foundation, as well as several major pharmaceutical companies.

In April, Plenge and his colleagues shared a description of how the Sage RA challenge will progress in a letter published online in Nature Genetics.

As it moves forward starting this fall, the challenge will have two phases, according to Plenge. In the first phase, participating teams will use existing genomic data sets — including a main set from the GWAS research conducted by Plenge and his colleagues — as well as new data collected specifically for the challenge.

"We have our GWAS data on about 2,700 RA patients treated with anti-TNF therapy, as well as what we've gone on to do, but have not published yet with polygenic modeling to show there is some signal there," said Plenge.

Additional RNA-seq, HapMap cell line, and some "curated" gene expression data will also be integrated, he said.

"We have agreements in place with all the people who have contributed data, and then a memorandum people sign to get access to the data,” Plenge explained. “The agreement is to use it for the purposes of the study and not beyond, so once the study is done that data will be returned."

The challenge, with its data-sharing framework, encourages collaboration and competition. As groups move forward with developing and testing their predictors, they have the option of collaborating through Sage's Synapse system to check in on each others' algorithms and further refine their own.

In the second phase, the group, led by Sage, will test the predictors in a new GWAS dataset of 1,100 patients provided by the Consortium of Rheumatology Researchers of North America and the Pharmacogenomics Research Network.

According to the challenge team, CORRONA is providing unique clinical samples from a prospective study of rheumatoid arthritis patients treated with anti-TNF therapy, and the PGRN is supporting the genotyping of these samples to provide a GWAS dataset for the validation phase of the challenge.

Teams will submit code, and Sage and the DREAM team will test all the predictors in Synapse against the validation dataset to determine accuracy.

According to Plenge, the group anticipates that the validation should be finished some time in early 2014, when the team with the most highly-predictive model will be the challenge "winner."

Plenge said that the list of participants is not set in stone yet but that 20-30 have indicated their interest in competing, and the Brigham and Women's team hopes that many more will end up joining.

While the main goal of the challenge is to try to identify a clinically useful predictor of anti-TNF therapy response, Plenge said that the effort will also have other measures of success.

"I'm confident that we will come up with a good predictor, but how clinically actionable it is going to be is less clear to me," he said. "At the very least, with larger datasets and [a] larger community we will do better than any one lab could do on their own."

"But the other important thing is that we are beginning a process," Plenge said. "I'd like to think of this as challenge number one of multiple future challenges. And as the process is iterated, I'm confident that over many challenges we will come up with a reliable predictor of therapy response."

Additionally, he said, the group hopes to apply the best predictor that comes out of the challenge to a citizen-science style clinical trial, using the Arthritis Foundation's Arthritis Internet Registry to further test the predictor's ability by tracking patients through self-reports and other measures.

If the challenge yields a novel response predictor for anti-TNF drugs that is ultimately commercialized, it would enter a sparsely populated RA molecular diagnostic market.

Crescendo Bioscience markets a multi-protein marker blood test, called Vectra DA that gauges disease activity in rheumatoid arthritis patients. The company has conducted a number of studies showing that the test can also track disease activity in RA patients treated with a variety of drugs, including anti-TNF therapies (PGx Reporter 11/9/2011).

In addition, Quest Diagnostics recently launched two molecular tests that the company is marketing as an aid for RA diagnoses. Both tests gauge the 14-3-3eta protein biomarker, which Quest has said might potentially be useful in gauging best responders to anti-TNF drugs (PM 4/12/2013).