AWS Public Sector Blog
Using artificial intelligence and machine learning to advance medical research
Academic medical centers (AMCs) are under pressure to reduce costs, innovate at scale, and improve operational performance. To do this, they’re turning to the cloud.
The cloud provides the availability, scalability, and security that medical researchers need to deal with large datasets that contain sensitive patient data to discover insights and deliver better care.
Two Amazon Web Services (AWS) Partner Network (APN) Public Sector Partners used the cloud to create solutions for AMCs that use large datasets to help advance medical research and analyze genomic data. Learn how these two partners are building solutions in the cloud to help AMCs further their mission:
Data insights from simulated patient data
When University of Texas Health Science Center at Houston (UTHealth) came to APN Partner Virtusa Corporation to find out how they can make the most of their data to benefit patients with sub arachnoid hemorrhage, they started with data analytics, artificial intelligence (AI), and machine learning (ML).
Virtusa turned to the cloud to design a new solution with Cardinal Health for UTHealth that creates simulated electronic health records (EHRs) and analyzes this information using AI and ML.
The solution, vLife, is a cloud-based healthcare and life sciences marketplace running on a HIPAA-compliant data lake, multiple data sources, pre-built APIs, AI, and ML models to uncover trends in patient treatment and outcomes. vLife can be applied to datasets and pull uncovered insights for specific illnesses, using AWS services like Amazon SageMaker, AWS Lambda, and Amazon Simple Storage Service (Amazon S3).
Virtusa and Cardinal Health will simulate the ML models on a data lake with EHRs for more than 30,000 patients. Based on the outcomes, UTHealth will use the data to tailor their own ML models in a team effort involving clinicians, computer scientists, and researchers with the UT School of Public Health.
vLife supports UTHealth’s efforts to identify common risk factors and the best possible treatment strategies after the patient understands their diagnosis. By applying their ML models to the test environment, UTHealth researchers can validate predictive models and ML algorithms before applying it to real patients and their care.
“Healthcare data is imperfect, and you’re helping close some of those gaps to increase the probability of findings for healthcare providers with vLife. For UTHealth, our endeavor have been in providing them simulated EMR data to help them test and validate their predictive models across multiple disease states,” said Kartik Iyengar, CTO of life sciences at Virtusa.
Learn more about Virtusa’s vLife and check out some of their available datasets on the Amazon Data Exchange.
Analyzing genomic workloads
In genomics, the cloud helps derive actionable insights from large, complex data, and scale up and down as needed. Goldfinch Bio needed help analyzing genomic data to help identify new therapeutics to treat kidney disease. APN Premier Consulting Partner Privo IT worked to transform a popular open source genomics tool originally developed for the Broad Institute, a joint venture between Massachusetts Institute of Technology (MIT) and Harvard University focused on genomics, to repackage the tooling on AWS. They created Hail on Amazon EMR.
Hail is an open source solution for working on secondary analysis of genomic workloads. Hail is designed to scale, so that genomics researchers can sift through the data and make progress on research quickly. The newly designed Hail on Amazon EMR runs on top of robust Amazon EMR infrastructure and leverages Amazon SageMaker Notebook instances, which are remotely connected to the main cluster via Apache Livy. An automated build process backed by AWS CodeBuild creates the custom machine images. Hail is designed to scale for multi-petabyte data sets and empower researchers to analyze variants across thousands of genomes painlessly.
With Hail on Amazon EMR, Goldfinch Bio saw reduced cluster deployment time, easier to manage scaling events, and more resilient workspaces (in the form of Sagemaker Notebooks), as well as saved time on operational cycles to maintain the infrastructure. “The benefit to AMCs is enabling scientists to focus on doing the actual science. With Hail on Amazon EMR, you’re empowering researchers by allowing them to focus on what matters—analyzing data, iterating quickly, and accelerating innovation,” Jack O’Brien, Director of Engineering, Privo.
Learn more about Hail and read the Goldfinch Bio case study.
Learn more about public sector healthcare, cloud computing for education, and research and technical computing on AWS.