Trace Id is missing
October 16, 2020

Adaptive Biotechnologies transforms data stream into new immune medicine platform with Azure

Founded in 2009, Adaptive Biotechnologies is a commercial-stage biotechnology company focused on harnessing the inherent biology of the adaptive immune system to transform the diagnosis and treatment of disease. Adaptive was built on the premise that the adaptive immune system can detect and treat most diseases in the exact same way, but the inability to understand precisely how that system works has prevented the medical community from fully leveraging its capabilities.

Adaptive Biotechnologies

Building on the foundation of its immunosequencing technology, Adaptive built a proprietary immune medicine platform over the last decade that is uniquely capable of decoding the genetic language of the adaptive immune system at scale to understand exactly how it works. Adaptive needed to synthesize this huge system of biology and tap into the full value of the massive clinical immunomics database generated, so the company turned to Microsoft Azure for compute, storage, and machine learning capabilities. With the power and scalability of this platform, Adaptive is poised to launch a new line of diagnostic offerings for the early detection of many diseases based on the immune response, and it is set to scale much more quickly.

“We use Azure cloud computing resources and machine learning capabilities to power our immune medicine platform, so we can rapidly map the adaptive immune system to the many different diseases it recognizes. Armed with this map, we can develop novel diagnostics for diseases, including COVID-19, and fuel the next generation of diagnostics.”

Mark Adams, Chief Technical Officer, Adaptive Biotechnologies

Power of the human immune system 

The immune system is astonishingly brilliant in its ability to precisely detect and attack disease and to record its activity throughout our lives. It accomplishes this feat through a massively diverse set of immune receptors on the surface of specialized cells. Adaptive Biotechnologies, based in Seattle, Washington, was founded on the premise that tapping into this information may enable a transformation in the diagnosis and treatment of disease. The commercial-stage biotech company helps researchers and clinicians gain a deeper understanding of the human immune system by sequencing the genomes of immune cells to read the immune systems of individuals and across the population.

The key cells of the adaptive immune system that enable our body to mount responses against antigens are called T cells and B cells. Each of these cells has a unique receptor (T-cell receptors and B-cell receptors, or TCRs and BCRs) that can recognize one or a small number of the millions of antigens to which our bodies are continuously exposed. Adaptive has built a platform that sequences these receptors, maps the TCRs to the antigens to which they bind, and can characterize TCRs and BCRs to identify the ones that are the most promising therapeutically. 

This innovative approach shows a lot of promise—and generates a lot of data. In fact, Adaptive’s dynamic clinical immunomics database includes more than 47 billion immune receptors to date. Adaptive determined that it needed to use high-scale compute resources and machine learning capabilities to unlock the full potential of the research data. “Thanks to AI and other novel technologies, today we can harness data to crack the code of disease and the immune response to disease in ways unimaginable until now,” says Julie Rubinstein, President of Adaptive Biotechnologies.

Expanded possibilities in the cloud

Adaptive worked with Microsoft to explore the cloud and create a roadmap for the company’s technology needs. Adaptive adopted Microsoft Azure to apply machine learning to exponentially accelerate the company’s ability to apply its proprietary immune medicine platform to gain novel insights from its clinical immunomics database. With a scalable immune medicine platform, researchers could begin computationally mapping trillions of TCRs to millions of disease-specific antigens that they are specifically targeted to attach to—called the TCR-Antigen Map—potentially enabling new approaches to diagnosing disease more precisely and earlier than is currently possible for many diseases. 

When researchers or clinicians submit samples for analysis, Adaptive uses the platform for end-to-end support, from sample intake to processing, to generate the sequencing data targeted to the immune response of a particular genome. It then sends that data through an automated in-house bioinformatics pipeline running on virtual machines created using Azure Virtual Machines. Adaptive researchers then explore the processed data and data science workbench developed using Azure components, including Azure Machine Learning and Azure Kubernetes Service (AKS), to train and evaluate models.

Adaptive uses this workbench to speed the development and deployment of these high-scale, reproducible machine learning experiments and to accelerate all its exploratory research. It stores its many terabytes of source and experimental data in Azure Blob storage. Then, the company takes advantage of virtual machines and Azure Application Gateway to provide web-based access to data and interactive research environments—both for clinicians to request reports on their patients and for researchers to see data results and use visualization tools to analyze the data.  

“We needed the scaling and flexibility in Azure to realize the full potential of our immune medicine platform and to develop the TCR-Antigen Map,” says Terry Franklin, Director of Production IT at Adaptive Biotechnologies.

Peace of mind for patients

In addition to using Azure to build out this new line of innovative diagnostics, Adaptive also uses the Azure platform to power laboratory workflows for existing products, including FDA-cleared clonoSEQ for the detection and monitoring of minimal residual disease for patients with certain blood cancers. By using this approach to training machine learning models, Adaptive invested a little bit of time upfront for the automation, but ultimately it made data processing and exploration significantly faster. 

Now, the company can process datasets simultaneously, and the accuracy of its results will be continuously improved and updated online in real time as Adaptive sequences more samples. Faster results are important to the quality of care and peace of mind for patients who are waiting for life-changing news, such as a cancer test result or new treatment for a disease. Moreover, by developing a cloud-first workbench, Adaptive can run experiments that are highly reproducible and deeply collaborative, so its geographically distributed data science teams can rapidly explore the data and develop and train new models.

“Using Azure, we can generate test results for patients much more quickly than we were able to do previously,” says Franklin. “When we were limited to our in-house hardware, we had jobs that sat waiting to be processed until the previous ones were done. Now, we can immediately and dynamically have as many systems as we need to match our lab’s output, generating lab results in parallel.”

Adds Mark Adams, Chief Technical Officer at Adaptive Biotechnologies, “With our Azure-based platform, we have been able to cut hours off the pipeline processing time.”

Transforming offerings and research

Since implementing Azure, Adaptive has transformed its ability to process and analyze massive amounts of data on a whole new scale. The company has fundamentally increased its productivity, but it’s not simply doing faster research and processing a higher volume—it is finding results that it could not even look for before. By combining Azure scalability and machine learning in its immune medicine platform, Adaptive is turning its existing data stream into a new diagnostics business.

For infectious diseases such as Lyme disease, a patient’s immune system may use thousands of TCRs to fight the pathogen—out of billions of unique receptor sequences that are on standby to fight other diseases. Although sequencing TCRs from the hundreds of thousands of T cells present in a small blood sample has been core to Adaptive’s business, the company now uses the cloud-scale machine learning and higher throughput sequencing to find patterns of TCRs that are shared across patients and map them to specific antigens.

For example, with this innovative approach, Adaptive identified shared T-cell signatures commonly found in Lyme disease. Within months of finding a particular signature, Adaptive launched a clinical validation study to assess the use of this signature as a diagnostic.

“With Azure Machine Learning, we can take a ton of data and train models to help us sort through that data in an efficient and effective way—especially if we don’t know exactly what to look for,” says Adams. 

Innovative approach to COVID-19

When cases of COVID-19 began to rise in the United States in early 2020, Adaptive and Microsoft extended their existing relationship to map population-wide adaptive immune responses to diseases at scale to identify how the immune system responds to COVID-19. Within weeks, Adaptive processed 500 million TCR sequences, using 29 compute-years, to identify TCR signatures of infection. As a result, Adaptive demonstrated that the T-cell signature it discovered could enable detection of SARS-CoV-2 infection. In a head-to-head comparison with two leading serology tests, Adaptive recently demonstrated the potential of its T-cell-based approach to detect immune response to SARS-CoV-2 earlier, and in less severe cases, than tests that detect antibody response.

On the strength of these findings, the company has begun pursuing an Emergency Use Authorization from the FDA for what could become one of the world’s first T-cell-based diagnostics. In addition, the company launched immunoSEQ T-MAP COVID to support scientific research and vaccine development. Microsoft and Adaptive are also making their findings freely available to researchers around the world via ImmuneCODE, a public database with hundreds of millions of TCRs from more than 1,500 COVID-19 patients to date.

Concludes Adams, “We use Azure cloud computing resources and machine learning capabilities to power our immune medicine platform, so we can rapidly map the adaptive immune system to the many different diseases it recognizes. Armed with this map, we can develop novel diagnostics for diseases, including COVID-19, and fuel the next generation of diagnostics.”

Find out more about Adaptive Biotechnologies on Twitter, Facebook, and LinkedIn.

“We needed the scaling and flexibility in Azure to realize the full potential of our immune medicine platform and to develop the TCR-Antigen Map.”

Terry Franklin, Director of Production IT, Adaptive Biotechnologies

Take the next step

Fuel innovation with Microsoft

Talk to an expert about custom solutions

Let us help you create customized solutions and achieve your unique business goals.

Drive results with proven solutions

Achieve more with the products and solutions that helped our customers reach their goals.

Follow Microsoft