Agilent Improves the Performance of Its Genomics Software and Saves Time Using AWS

Established in 1999, Agilent is a life sciences company that provides software, instruments, and consumables for scientists. Its genomics division helps researchers unlock the potential of genomic data.
 
“The burden of data interpretation and reporting increases as labs expand their testing menu to include more genes, increasing the number of variants that need to be analyzed—making it critical to be able to prioritize clinically relevant variants that might play a role in the patient’s phenotype or tumor type,” says Elias Hage, product manager of tertiary analysis software solutions at Agilent. “Our software helps clinical genetics and molecular-pathology labs reduce turnaround time and increase productivity through standardization and automation of triage, review, and classification of genomic data.”
 
The Alissa Clinical Informatics Platform has three components: Alissa Reporter, which analyzes raw sequencing data and identifies variants; Alissa Interpret, which interprets and annotates the variants; and Alissa Annotation Server, which compiles annotation sources.

From the beginning, Alissa Interpret and Alissa Annotation Server have run on AWS. To improve performance and scalability, Agilent migrated Alissa Reporter from a third-party data center to AWS in 2018. It engaged AWS Enterprise Support, which provides 24-7 technical support from high-quality engineers, to optimize the process. “We started using AWS from the very beginning because our workloads needed massive amounts of compute power and storage,” says Gert Thijs, product manager of Alissa Annotation Server. “AWS has given us the needed flexibility and has made our solution platform agnostic.”

To run large-scale genome analyses, Agilent uses Amazon Elastic Compute Cloud (Amazon EC2), which provides secure and resizable compute capacity for virtually any workload. Agilent can select the best available compute instances for a certain computation using AWS Auto Scaling, which automates application scaling to optimize performance and costs. Amazon CloudWatch, a service that collects and visualizes near-real-time logs, metrics, and event data in automated dashboards, is used to observe and monitor Agilent’s resources and applications on AWS.

Agilent implemented hardware-optimized variant-calling algorithms developed by NVIDIA Clara Parabricks that require GPUs instead of regular CPUs. To improve processing speeds for its variant-calling workflows, the company adopted Amazon EC2 G4 Instances, the industry’s most cost-effective GPU instances for machine learning (ML) inference and graphics-intensive applications, and Amazon EC2 G5 Instances, the latest generation of NVIDIA GPU-based instances. Using AWS Auto Scaling groups, Agilent can define the most appropriate instances for a certain computation. With this implementation, Alissa Reporter can complete a secondary analysis (the mapping a calling step) in 9 minutes instead of 9 hours—a 96 percent reduction—and run hundreds of workloads in parallel.

Amazon CloudWatch collects and visualizes near-real-time logs, metrics, and event data in automated dashboards, which helps Agilent streamline infrastructure and application maintenance. “Customers are very happy with the decrease in turnaround time,” says Christina Hörnlein, product manager of Alissa Reporter at Agilent. “They’re pleased with the scalability as well. When you can just analyze samples in parallel, you don’t have to prioritize workloads.”

Outside of getting results faster, Agilent can also reduce costs for customers with per-second billing on AWS. “We have computations that can take hours, and some just minutes,” says Thijs. “This makes the per-second billing very convenient to scale automatically.”

Because Agilent’s customers work with sensitive health information, its systems must comply with HIPAA and ISO requirements. For example, the risk of inadvertent access must be low, and all data must be encrypted and shielded from the outside internet. On AWS, Agilent can configure frameworks that meet these regulations. “We have an AWS business agreement that demonstrates that our services are running in a secure environment,” says Thijs. “This is recognized by different authorities who certify our product and shows that we are compliant with clinical data-interpretation rules.”

The benefits extend to Alissa Interpret and Alissa Annotation Server. On AWS, Agilent hosts the different knowledge sources needed for annotation and interpretation, with about 1.5 TB of annotation data served to all connected applications. With flexible AWS architecture, Agilent can scale as its customers grow. “Our customers typically start with a few analyses but scale fast thanks to all of these automation capabilities,” says Thijs. “If we see a customer increase their analysis volumes, we can scale to grow with them without causing any service interruptions.”

Faster analysis and interpretation of genomic data are helping researchers suggest targeted therapies to doctors sooner. Scientists can broaden their research and recommend more treatments than they could previously. “On AWS, we make it possible for customers to get fast results,” says Joachim de Schrijver, product owner of Alissa Reporter. “If something takes 15 minutes instead of 8 hours, you can run 50 parameter sets in a single day, where it might take a month to get the same results. We can speed up development and pass on time and cost savings to the customer.”