Gut Instinct: Human Microbiome May Reveal COVID-19 Mysteries


Days sooner than a national lockdown within the U.S., Daniel McDonald realized his existence’s work had assign a selected tool in his fingers to fight COVID-19.

The assay kits his crew became once about to contain made by the tens of hundreds would be repurposed to serve perceive the unconventional coronavirus that causes the disease.

McDonald is scientific director of the American Gut Venture and the Microsetta Initiative, section of an emerging discipline that learn microbiomes, the collections of single-cell creatures that fabricate up grand if no longer most of existence in and around us. The assay kits had been the first in an effort to safely grab and ship samples from human feces preserved at room temperature.

The kits originally focused gargantuan learn in microbiology. Nonetheless McDonald and his colleagues knew they wished to pivot into the pandemic.

With cautious screening, samples would possibly perchance well presumably even point to patterns of how the mutating coronavirus became once spreading. That data would possibly perchance well presumably presumably be gold for public health consultants attempting to tedious the growth of unique infections.

The crew also hopes to gather staunch ample data from participants to let researchers explore one other mystery: Why does the virus fabricate some of us very ailing while others point to no indicators in any respect?

“All and sundry right here is in overall desirous about doing one thing that can well presumably even serve set lives,” acknowledged McDonald, section of the 50-person crew in Opt Knight’s lab on the College of California, San Diego.

“We are lucky to work shut and collaborate with consultants in RNA and assorted areas applicable for finding out this virus,” he added.

Hitting the Accelerator on the Ethical Time

Because the kits had been taking form, the neighborhood had one other stroke of correct fortune.

Igor Sfiligoi, lead scientific instrument developer on the San Diego Supercomputer Heart, ported to NVIDIA GPUs the most neatly-liked version of the crew’s performance-hungry UniFrac instrument, which is mature to match microbiomes. The effects had been aesthetic.

A genetic evaluation of 113,000 samples that would contain required 1,300 CPU core-hours on a cluster of servers (or about 900 hours on a single CPU) accomplished in lower than two hours on a single NVIDIA V100 Tensor Core GPU – a 500x speedup. A cluster of eight V100 GPUs would gash that to lower than 15 minutes.

The port also enabled particular person researchers to trudge the evaluation in 9 hours on a workstation equipped with an NVIDIA GeForce RTX 2080 Ti. And a smaller dataset that takes 13 hours on a server CPU now runs in staunch over one hour on a laptop laptop with an NVIDIA GTX 1050 GPU.

“That’s sport altering for of us that don’t contain procure admission to to high-performance laptop systems,” acknowledged McDonald. To illustrate, particular person researchers would have the opportunity to employ UniFrac as an growth of search tool for advert hoc queries, he acknowledged.

With the lab’s cluster of six V100 GPUs, it would possibly perchance per chance well presumably even launch to kind out evaluation of its rising datasets.

Sfiligoi’s work on 113,000 samples “arguably represents the largest evaluate of microbial existence to this point,” McDonald acknowledged. Nonetheless the lab already has a repository of about 300,000 public samples and “it acquired’t be grand longer sooner than we’re properly in grand more than one million samples,” he added.

GPUs Velocity Up UniFrac Three Methods

Three ways had been key to the speedups. OpenACC accelerated the many tight loops within the Striped UniFrac code, then Sfiligoi utilized memory optimizations. Downshifting from 64-bit to 32-bit floating-point math delivered extra speedups with out affecting the accuracy the experiments wished.

Sfiligoi of the San Diego Supercomputing Heart ported UniFrac to GPUs.

Sfiligoi accomplished the initial OpenACC port in a matter of days. Other optimizations came in incremental steps over a few weeks because the crew acquired a closer understanding of UniFrac’s compute and memory-procure admission to wants.

The work came on the heels of landmark effort Sfiligoi described in a session at GTC Digital. He became once section of a crew that harnessed exascale performance from GPUs on public cloud companies and products for learn in astronomy.

NVIDIA is participating with Sfiligoi on his subsequent job. He aims to integrate his GPU optimizations on UniFrac into the instrument microbiologists employ daily.

Knowledge Flood Would Swamp CPU-handiest Programs

Meanwhile, McDonald and his crew contain to adapt UniFrac to work with viral data. They also face heady challenges turning big amounts of files they would possibly be able to generate into properly organized and blunder-free datasets they would possibly be able to route of.

On the tech front, the neighborhood wants a entire bunch storage and compute performance. To compare what would possibly perchance well presumably even one day amount one million microbiomes would possibly perchance well presumably even require 20 petabytes of storage and greater than 100 million CPU cycles/three hundred and sixty five days.

“I’d love to explore different that pushed onto GPUs,” McDonald acknowledged.

The work has gargantuan doable given how long the clan of coronaviruses has been affecting every folk and farm animals.

“All and sundry on this planet has felt these impacts on productivity in some advance. Now we are in a position to delivery out to know greater arrange this family of viruses which were with us a truly long time,” he added.

The efforts in San Diego are section of a gargantuan network of learn projects leveraging NVIDIA GPUs and high performance computing to fight COVID-19.

More than 30 supercomputing companies and products worldwide spanning companies and products in Asia, Australia, Europe and the United States are engaged within the pronounce. The COVID-19 High Efficiency Computing Consortium on my own has greater than 30 entertaining projects with procure admission to to 420 petaflops of energy that capabilities 41,000 GPUs.

Image at top: Opt Knight (left) and Daniel McDonald within the usa Knight Lab. Photo courtesy Erik Jepsen/UC San Diego Publications

Leave a comment

Your email address will not be published.