Next-Generation Sequencing technologies are generating increasingly larger, more complex data, and traditional CPU processing is no longer able to keep up with the demands. Healthcare, biotech, and pharmaceutical firms are turning to GPUs to process the datasets now generated in genomic sequencing. As demand rises in a growing number of fields, GPU processing power is becoming necessary to meet that demand.

On April 20th, 2022, Hewlett Packard Enterprise and NVIDIA hosted Transforming Genomics with GPUs: Enabling Improved Outcomes in Life Sciences Research and Patient Care, a panel discussion of use cases and best practices in the biotech and genomic industry featuring thought leaders in the field.

Here are a few important takeaways from the conversation:

1.Keep your data organized

George Vacek of NVIDIA spoke about the importance of good data management when it comes to building and testing machine learning algorithms for processing large amounts of genomic data. “When you’re applying machine learning and deep learning on any data set, the important thing is the data you’re training it with and that it is well annotated.”

Without well-organized data, it’s difficult to understand what kind of results machine learning algorithms are generating and build accurate models from those results.

2. Make data accessible

“Who is the end user?” asked Dan Ryder , Ryder of Bridge Informatics while talking about data management and accesibility. “It’s typically very challenging for a biologist or an ex-bench scientist to use a command line interface.”

He emphasized the importance of making data accessible to the people who will be using it in a form they will be familiar with, which often means getting databases into excel or other interfaces that biologists can use.

3. Optimize your metadata

Genomic data is often made up of many small data files with large metadata files attached to it.Christopher Davidson of HPE proposed a strategy for optimizing genomic data processing, where metadata is processed on flash drives and object data is processed on disc drives.

“You’re speeding up the metadata-aspect of what you’re looking at,” Christopher said, “and that can have tremendous impact on looking at data on the analysis phase.”

4. Manage data storage

Data management can be an incredibly complex task, weighing prices between data storage, transfer, and archiving, which need to be planned carefully to avoid expensive mistakes. Beyond that, servers can often pose their own complications in the physical world.

Marty Poniatowski of HPE described a use-case in which a client of his was a marijuana grower with servers in the United States. “They got slapped on the wrist and had to get some storage on-site really quickly to adhere to data sovereignty.” Data management can be complicated, and it’s important to do thorough planning in order to manage project budgets.

5. The genomics sequencing industry is growing

Genomic sequencing is still a very new technology with new tests and panels being introduced to the market constantly. Christopher Davidson speculated on the growth of these technologies, and the ways in which they will be implemented in future healthcare interactions. He spoke about a vitamin d test that he had years ago which was then cutting-edge but is now commonplace.

“It’s maybe more of a grassroots movement,” Christopher said. “How are physicians and clinicians being taught or learning about this?”

Interested in learning more about genomic sequencing and GPUs? Access the full event content by filling out the form at the top of this page.