Amazon Web Services held an online panel discussion Thursday that looked at how the company's cloud infrastructure is supporting the COVID-19 response, from outbreak prediction to vaccine development.
The rapid progression from viral outbreak in China to full-blown global pandemic has magnified the role of clinical researchers, biotech companies and drug manufacturers in the global response to the virus. For the key players in this space, thehas become a high stakes data challenge and cloud technology case study.
AWS customers BlueDot, Lifebit, AbCellera, Moderna, UC San Diego Health System, and Babylon are using a range of cloud technologies to increase the pace of innovation, accelerate development timelines and help improve outcomes during the COVID-19 pandemic.
BlueDot is a Toronto-based company that is harvesting data from medical publications across the world and applying machine learning to that data to identify trends and signals related to the coronavirus pandemic. The aim is to leverage big data and analytics to quickly detect outbreaks around the world, anticipate how the outbreaks might spread, and also to anticipate the consequences that an outbreak could have.
BlueDot co-founder and CEO Kamran Khan explained that he's been focused on the impacts of emerging infectious diseases for most of his career, starting the with SARS-CoV outbreak in 2003. Khan is among the majority of global scientists and researchers highlighting the need for an early warning system for infectious diseases.
"One of the things we learned during the SARS outbreak is that if we wait for official reports from public health agencies, we may not always get that information in the most timely manner," Khan said. "Our data scientists and engineers and clinicians have been building a platform that is gathering information on over 150 diseases and syndromes across 65 languages, and collecting this information every 15 minutes, 24 hours a day. This is where natural language processing and machine learning comes into play."
Khan said BlueDot is using natural language processing (NLP) and machine learning powered by AWS to extract information on various pathogens, including location and time, and other contextual data such as case counts and deaths. The ultimate goal is to turn this unstructured text data into organized, structural spatiotemporal pathogen data, where the space and time and name of the pathogen becomes known.
"We used this platform to pick up news of an outbreak of pneumonia in Wuhan back on the morning of December 31 by translating an article in Chinese and having the machine present it to our team as a threat that we should be paying attention to," Khan said.
By early January, BlueDot had submitted their research to a medical journal highlighting the cities most at risk for viral spread and outbreaks due to the novel virus in Wuhan. The company remains focused on detecting future large outbreaks.
Lifebit is a partner of AWS that focuses on genomic analysis and research, and the impact that genomics can have on the success of clinical vaccine trials for COVID-19. LifeBit is working with the UK's National Health Service as its end-to-end research platform, enabling researchers to access, query and analyze genomic data and also combine it with their own data to run joint analysis as if the data was in one place.
"The common goal in all of this is to identify the genetic variances that can explain why some individuals get affected severely by COVID-19 where others get mild symptoms or are completely asymptomatic," said Maria Dunford, CEO of Lifebit.
AbCellera is another AWS partner focused on early drug discovery. The company uses blood samples from COVID-19 patients and labels the antibodies. From there it uses machine vision on AWS to look at cells and identify the kind of antibody that could be given to a COVID-19 patient as a treatment.
AbCellera CEO Carl Hansen said his company has built a full-stack technology platform over the last seven years that's working to discover new therapeutic antibodies for infectious diseases. The company uses a range of AWS technologies including Elastic Bean Stalk for hosting applications, RDS for database management, and Elastic Compute Cloud for server scaling.
"Drug discovery typically takes anywhere from 5 to 15 years and succeeds less than 5 percent of the time," said Hansen. "We need technology firms, not biotech firms, that are investing in solutions to make the process faster and to sharpen the ax of drug discovery. We believe if we can put the best science and the best innovators in touch with the best technologies we will get faster drugs to patients and everyone wins."
AbCellera's pandemic prevention platform was developed and refined over two years under DARPA's P3 program to tackle pandemic threats like COVID-19.
On the same day the company was preparing for its third pressure test of its platform, the first American patient in the US was identified. With that, the team shifted its focus from a simulated pandemic to the real thing. Once the company obtained its first blood sample from a recovered COVID-19 patient, it took 9 days to identify roughly 500 unique fully human antibodies that were subjected to analytics. Each antibody was evaluated against at least 500 different qualities to filter down 24 antibody frontrunners for drug development with pharmaceutical partners. On April 15, one antibody was selected and pharmaceutical firm Eli Lilly began manufacturing the LYCoV555 vaccine. Human trials began on June 1.
Moderna is using AWS compute to identify protein families that can be used as part of the COVID-19 vaccine. Moderna chief executive Stephane Bancel described how the biotech company has ramped up development of the SARS-CoV-2 vaccine, from the sequencing of mRNA-1273 in January to several phase 3 clinical trials planned for this month.
"If you look at SARS, it took 20 months to go from sequence of a virus to something in clinical study," said Bancel. "It took us 63 days."
- AWS touts current way of working as the new normal
- Will the COVID-19 pandemic have a lasting impact on cities?
- Google expects its staff to work from home until 2021 and it's not alone
The UC San Diego Health System applied machine learning to chest x-rays to spot early signs of COVID-19 infection. Researchers also applied machine learning to identify which x-ray films were likely to represent COVID-19 infections and should be viewed more quickly by the technicians to accelerate time to treatment.
Mike Hogarth, clinical research information officer for the UC San Diego Health System, explained how his team began using an algorithm to spot subtle changes to chest x-rays that would suggest the need for a CT scan. Early on in the COVID-19 pandemic, doctors on his team applied this algorithm to chest x-rays that were published by other clinicians. The algorithm was successful in detecting normal and abnormal chest x-rays and eventually became the core component of AI-enabled image analysis that's processing chest x-rays of potential coronavirus patients.
Babylon is a company focused on the telehealth market. Its platform applies machine learning to language to allow patients to have a natural language conversation with health providers on a range of coronavirus related questions. The company recently developed the Babylon COVID-19 Care Assistant, which aims to help free up clinical resources while also providing patients with appropriate care based on their individual needs.
In his presentation, Babylon CEO Ali Parsa stressed how these types of virtual services are critical to making healthcare more accessible and less costly, particularly in times of a pandemic.
"It is time to change the way we've been delivering healthcare," said Parsa. "It is possible thanks to organizations like AWS, for making cloud so widely available, and to the mobile networks, and to the other technology companies out there that have provided so much of the infrastructure that we need. It is time to look at healthcare in a fresh way."