DKFZ tested their first Convey hybrid-core computer in the summer of 2012, using the server to accelerate a portion of their next-generation pipeline. Results were extremely positive; one of their initial jobs went from over 18 hours on an 8 core system to a mere 48 minutes on the Convey server.
Based on that success, DKFZ purchased two Convey HC-2ex servers a few months later. These hybrid-core systems were integrated into an existing 1,600 compute core cluster. The Convey systems are now routinely used at DKFZ in the production analysis for all BWA algorithms.
"We are extremely pleased with the Convey systems. Depending on the data, we are typically seeing between 10-20x acceleration, which is substantial", explained Dr. Benedikt Brors, Group Leader Computational Oncology, Division Theoretical Bioinformatics at DKFZ. "And our entire work flow from one end to the other is reduced by a factor of two."
Convey's hybrid-core architecture pairs classic Intel x86 microprocessors with a coprocessor comprised of field-programmable gate arrays (FPGAs). Particular algorithms - BWA-based alignment, for example - are optimized and translated into code that's loadable onto the coprocessor at runtime. The result is order of magnitude performance increases with less power, space and cooling requirements.
The Heidelberg DKFZ facility currently processes 5 terabytes of data per day conducting NGS analysis. Prior to the Convey installation, processing massive amounts of data in a timely fashion was a challenge for DKFZ - both from a resource and time perspective.
One of the most time-consuming steps of analyzing a patient's genome involves using the Burrows-Wheeler Aligner (BWA) - the preferred algorithm for fast, accurate alignment to a reference genome of the short reads typical of high-throughput sequencers (HTS). BWA was taking up to 80% of CPU time in the data centre's pipeline.
"Accelerating this portion of our pipeline allows us to get answers substantially faster", noted Dr. Benedikt Brors. "In addition, saving processing time reduces power consumption and cooling - which is a huge advantage in saving money and a step forward in our commitment to pursue green technologies."
This DKFZ effort is part of the International Cancer Genome Consortium (ICGC) to study tumours from more than fifty types of cancers. Their goal is to develop better ways of diagnosing, treating and preventing many types of cancer. The scientists at DKFZ are currently participating in three different ICGC cancer research projects: paediatric brain tumours, malignant lymphoma, and prostate cancer. Their research focuses on mapping how cancer changes the genomes of cells and causes mutations within normal biological pathways.