Unbeknownst to most people, scientists at some of the world’s leading companies have developed a powerful tool that could speed up genome sequencing by orders of magnitude at a cost that is a fraction of what it used to be just a years ago. That tool goes by the name GOAST which stands for Genomics Optimization and Scalability Tool. Developed by Lenovo, a China-based company better known for its consumer electronics in partnership with Intel of the US, GOAST bundles hardware and software and is powered by high performance computing and AI.
Rolled out in 2019, GOAST has been used to decode and sequence genomes at up to 167 times faster than any other computing technology currently available. What used to take months and months to map a single genome has recently taken only 48 minutes. Beyond genomics, the tool is also being used to develop more climate-resistant crops to improve yield and food security. It is also being used in health sciences to test an individual’s susceptibility to diseases and predict responses to a specific treatment to eliminate unnecessary side effects.
In addition to reducing the amount of time used to map a single genome, lowering costs was a key consideration when developing GOAST, says Sumir Bhatia, president of Lenovo’s Infrastructure Solution Group in Asia Pacific (APAC). “[In line with] our mantra of smarter technology for all, GOAST is about democratising knowledge and really lowering the barriers for usability … Since GOAST costs about a third less than previous and boutique solutions, smaller research organisations and the ancillary research organisations that are also doing research for bigger labs are able to afford it.”
Thinking Like Scientists
Developing GOAST was not an easy feat. First, there were mixed variables and the team at Lenovo had to match them correctly using the right tools, software, utility and hardware. “There were variables like 30 single threaded non-distributed and distributed tools woven and interwoven throughout the workflow,” says Bhatia. Besides that, Lenovo had to ensure that GOAST is a plug-and-play tool. “Research organisations shouldn’t be spending time configuring — or tuning — their hardware and software. They should just have to put the tool in to work and get the results.”
Dr. Mileidy Giraldo, Global Lead, Life Sciences Lenovo HPC & AI, said the key to success was approaching the challenge of speeding up workloads through a scientific lens. “Like any good scientist, we wanted to build on Intel’s work. So we started with the Intel Select Solutions for Genomics Analytics reference architecture and took a wider approach to seeing what additional performance we could squeeze out of standard components,” she explained.
When Lenovo started, most researchers expected that speeding up genomics workloads would mostly be a matter of adding more memory. But instead of accepting the status quo, the Lenovo team took a systematic approach to exploring other options. They started with testing the lowest amount of memory that would suffice for a system and worked their way up to the highest possible amounts and let the data speak.
Lenovo also explored the performance impacts of changing system variables, including software parameters, the gamut of Intel Xeon® CPU SKUs and other Intel HPC building blocks. “One of our biggest findings was that a high number of cores are the most important feature for genomics and the other ‘omics’ disciplines. The other big lesson was the importance of learning how to optimize scripts and tools to the hardware.”
Dr. Giraldo said that Intel’s guidance along the way was invaluable. For example, Intel Life Sciences engineers shared knowledge about workloads and provided important resources, such as the optimal scripts and data sets for them. Intel also trained the Lenovo team on how to size and scale the genomics architecture from one node all the way up to cluster-sized based on customer needs. “Intel gave us a head start, and Lenovo took things farther and figured out how to address the complexity of that sizing for workloads like transcriptomics and proteomics,” she explained.
“Thanks to our internal work and the collaboration with Intel, at this point we have a clear understanding of how to dial the reference architecture up or down based on specific customer needs around performance versus budget.”
That said, Bhatia asserts that the tool is “not the end”. There are always further improvements to be made with the next level of infrastructure, stronger central processing units (CPUs) to improve GOAST’s effectiveness while further reducing the cost. He adds: “We’re now looking at how we can make this tool agile and quicker. It could sequence a genome in 48 minutes today, but perhaps that time could be halved tomorrow. That’s always the objective, to keep improving the tool.”
A tech talk by Dr. Mileidy Giraldo, Global Lead for Life Sciences at Lenovo HPC & AI, illustrating how Lenovo’s revolutionary Bioinformatics-optimized architecture, an Intel Select Solution, is driving new hope for scientists and clinicians with unprecedented performance and efficiency.