A Vancouver startup that detects deadly pathogens in the air before they strike humans is getting a shot in the arm from IBM Corp.’s cloud and machine learning technologies.
The OneTest platform from Fusion Genomics is designed to spot contagious diseases like Zika virus, avian flu and swine flu while they’re still airborne but haven’t infected humans yet.
“Our focus at the moment is the undiagnosed infection specifically,” said Mohammed Qadir, president and chief scientific officer of Fusion Genomics.
OneTest takes air samples from areas where there’s a high risk of infectious disease. The platform’s software maps out the DNA sequence of any pathogens in the air sample. Using bioinformatics and machine learning (ML), the tool compares the sample DNA with Fusion’s proprietary database of over one million infectious disease genomes.
“This is the genome (database) of every known infection. This procedure is what is known as target capture sequencing,” Qadir said.
If OneTest detects a deadly contagion in the air, it generates a predictive model of how likely the pathogen is to cause harm to humans. Armed with that information, agencies like hospitals or public health officials can take steps to prevent a full-blown epidemic before it starts.
“They get the analysis feedback and based on the expertise of experts who review the data, they can then issue warnings,” Qadir said.
Preventing pandemics
Health officials in London, Ont. issued a warning last week about an outbreak of Group A streptococcus that has killed nine people and sent 30 more to intensive care over the past 18 months. Strep A, which can cause toxic shock and flesh-eating disease, is spread through contact with infected wounds, dirty needles or nose and throat secretions. Half of the 132 London patients lived on the street or in homeless shelters.
While OneTest could be used in settings like shelters, hospitals, prisons and refugee camps, Qadir said Fusion has completed “a very small trial installed inside pig farms. We were able to (detect) the pig virus from those aerosolized (sample) containers.”
Swine flu can be fatal in humans. In India alone, it has killed 8,543 people since 2010.
The entire OneTest process, from assaying the air to determining human risk, can be performed on-site in a high-risk environment and takes 14 to 72 hours, Qadir said.
ML and cloud
Cloud and machine learning underpin the whole OneTest solution, said Greg Stazyk, chief technology officer at Fusion Genomics.
“We use IaaS (infrastructure-as-a-service). It allows us to do our ML algorithms and data processing using the best resources tailored to the specific need, on-demand and on a scale that wasn’t possible before. It’s very batch intensive. There’s a lot of analysis that needs to be done at a time, so we don’t want those resources sticking around when we don’t need them. IaaS allows us to scale up and down dramatically,” Stazyk said.
IBM provides the cloud services and framework. That includes the nuts and bolts of FusionCloud, a private cloud portal where Fusion Genomics customers can upload, analyze and share data. Fusion gains quicker, easier access to computing power and security safeguards through IBM than it could ever build on its own, Stazyk said.
“There’s a lot of (IBM) security and network architecture and the ability to overlay our own requirements on top of that. There’s also their global presence, storage and computational resources in geographical regions where, for regulatory reasons, we want certain clinical data to be housed and stored within that region. A prime example is Canada.”
Unlike hospitals that collect and analyze heavy data volumes 24 hours a day, the medical data processed by Fusion Genomics could fluctuate periodically or, in the case of something like common respiratory flu, seasonally. So it needs a cloud with security that is strong yet flexible as its needs change, said Allen Lalonde, senior innovation executive at IBM Canada.
“It’s not just secure but scalable with that security. It allows that security to be scaled up or down depending on the needs or wants it’s being applied to,” said Lalonde, who’s also director of the Toronto-based IBM Canada Research and Development Centre.