NVIDIA Accelerated Computing Solutions Reduce Data Processing Time by up to 30x
NVIDIA announced the installation of an NVIDIA DGX™ A100 system and NVIDIA Clara Parabricks sequencing analysis software at the National Biobank of Thailand (NBT) to accelerate genomic sequencing as part of the government’s plan to promote genomic medicine in Thailand.
In 2019, the research institution was entrusted to design and implement the IT infrastructure for a national project, called Genomics Thailand (GeTH), aiming to introduce genomic medicine as a common medical service. One of the GeTH’s flagship projects involves extracting individual’s genetic variations from 50,000 Thai volunteers’ whole genome sequencing (GeTH50K) data. This work will provide a new collection of variants that is better to represent the Thai population than those from publicly available databases. Thus, such information is a good genetic reference that helps identify causative mutations.
GeTH 50K database harbors variants distributed across the entire human genome which are extremely useful in population genetics. Those rare variants from the database may have some medical importance.
NVIDIA Clara Parabricks: Getting Started
Massive Amount of Data to Process
The biggest challenge in this project is the sheer size of whole genome sequencing (WGS) data. The sequence data for an individual’s genome contains more than 100GB that must be sequentially aligned to a human genome reference to identify potential variants of the individual. This process results in an extra 100GB in total of over 200GB per sample. The parallel processing power of GPUs dramatically accelerates the entire process.
Identification of variant, called variant calling is the required process in genomics medicine. Accurate and rapid processing of WGS data makes it possible for patients to be treated with precise and personalized care, improving quality of life by reducing hospital visits and the associated costs.
Power and Computational Framework
To accelerate the variant calling process from WGS data, NBT utilizes NVIDIA DGX A100, the universal system for AI workloads. NVIDIA DGX A100 integrates eight NVIDIA A100 Tensor Core GPU accelerators delivering 5 petaflops of AI performance for researchers. This unprecedented compute density, performance and flexibility enables NBT to consolidate training, inference, and analytics into a unified, easy-to-deploy AI infrastructure.
On top of the accelerated hardware, NBT uses the NVIDIA Clara Parabricks computational pipelines, which support several genomics applications. Employing NVIDIA’s CUDA®, HPC, AI and data analytics stacks, Clara Parabricks Pipelines empower researchers to build GPU-accelerated libraries, pipelines, and reference application workflows for primary, secondary and tertiary analysis. The complete portfolio of off-the-shelf solutions is coupled with a toolkit to support new application development to address the rapidly evolving needs of high throughput genomic labs.
Accelerated Solution Expedites Discoveries
“The accelerated solution from NVIDIA lets us carry out the WGS variant discovery process efficiently and with high confidence. By pairing NVIDIA DGX A100 with NVIDIA Clara Parabricks, we have been able to reduce our WGS data processing by four months. Processing time per individual user has also been shortened from more than 30 hours to just one to two hours,” said Sissades Tongsima, Director of NBT. In addition, NBT’s A100 platform can be used for other NBT’s AI-related tasks.