Interesting Applications of AI in Covid-19 Research
On November 17th of 2019, the first case of Covid-19 was confirmed in China. Since the initial outbreak, the virus has rapidly spread across the world with 9.8 million confirmed cases worldwide (note it is currently June 26th ) and many countries have gone into lockdown. With the world’s attention focused on Covid-19, there have been thousands of research papers and abstracts written on the issue of Covid-19.
On March 14, the White House Office of Science and Technology Policy worked in conjunction with several other organizations in order to release the Covid-19 Open Research Dataset which was geared toward data mining and natural language processing. In this data set there are currently around 140,000 different papers on Covid-19 all formatted in a way that is easier for algorithms to parse. Every week, the White house asks a question and offers a 1,000 dollar prize for the team with the best answer. This is all done on Kaggle which is a subsidiary of google which has a community of data scientists and machine learning experts. Furthermore, many companies have released search engines such as the Covid-19 Research Explorer by Google. It allows people to ask questions and it returns articles with the relevant parts highlighted. After seeing these areas where AI could be applied to assist Covid-19 research, I got curious and decided to look into it more. These are a couple of the research papers and projects that I found were really interesting.
Potential non-covalent SARS-CoV-2 3C-like protease inhibitors designed using generative deep learning approaches and reviewed by human medicinal chemist in virtual reality.
This is a preprint article by Insilico and noname.ai. The article proposes 10 different potential small molecule inhibitors which is essentially a molecule that inhibits or prevents a target molecule from latching onto another molecule. These 10 different inhibitors were designed by several machine learning algorithms. The machine learning algorithm essentially analyzed the area of the Covid-19 virus that is used to latch onto your cells or the binding site. Then based on the binding site, the algorithm would design a potential molecule that could inhibit the Covid-19 virus. The molecules could then be reviewed within a virtual space by a chemist by a program made by noname.ai.
2. Artificial intelligence–enabled rapid diagnosis of patients with COVID-19
In a research paper released by Mount Sinai hospital, the researchers designed and trained a neural network that to at the radiology data from a CT scan and clinical data to determine whether a patient had Covid-19. In a test of 25 positive cases, the algorithm determined that 17 out of 25 of them were positive while a senior thoracic radiology fellow and a thoracic radiology fellow determined that 0 out of 25 of the cases were positive. This algorithm demonstrates large amounts of potential and could be used in hospitals across the world.
3. Watson AI Lab at MIT
The Watson AI lab decided that it would fund 10 different research projects that were geared to “addressing the health and economic consequences of the pandemic”. Here are a couple ones that were pretty interesting.
a) Returning to normal via targeted lockdowns, personalized treatments, and mass testing
This project looks at the effects of lockdowns and other measures that are meant to “flatten the curve”. Then the researchers will use machine learning models to identify which patients are the most vulnerable to the Covid-19
b) A Privacy-first Approach to Automated Contact Tracing
This project identified that there is a privacy issue with using smartphone data to identify if you were in contact with a person with Covid-19. This study plans on using encrypted bluetooth data to make sure that private information remains private and secure.
This is my first ever blog post, please leave a comment with some feedback, if you liked it share it with a colleague or maybe start a discussion in the forums. Thanks for reading!