Artificial Intelligence May Help Develop COVID-19 Treatments, Tests

Throughout the COVID-19 pandemic, artificial intelligence tools have played a significant role in tracking, controlling, and predicting the spread of the virus.

For more coronavirus updates, visit our resource page, updated twice daily by Xtelligent Healthcare Media.

Now, researchers are leveraging artificial intelligence tools to develop more effective treatments and testing for coronavirus.

A team from the University of Washington Bothell used deep learning to build a software tool that could help design vaccines.

Called DeepTracer, the tool is able to analyze a three-dimensional image of a virus protein molecule and trace the connections of its atoms.

“If you know the actual atomic structure of the viral protein, you will know how to speed up the development of vaccines or drugs,” said Dong Si, an assistant professor in the School of STEM’s Division of Computing & Software Systems. 

The virus images come from electron microscopes, recorded under cold cryogenic conditions. Cryogenic electron microscopy (cryo-EM) reduces electron irradiation damage to biological samples by freezing the specimens. Different images can be combined for a 3D reconstruction, called 3D electron microscopy or 3DEM. But the results don’t show everything scientists need to see, which is why the images have to be processed further.

Most researchers in the field use physics or traditional computational methods to predict atomic structure, but Si and his team formed a data mining and deep learning approach to refine predictive models.

“We are one of the first groups in the world to use deep learning to do this three-dimensional EM atomic protein complex structure prediction,” said Si.

DeepTracer has demonstrated its potential to enhance the field of cryo-EM.

“It fills the gap between cryo-EM maps at high resolution and atomic modelling from sequence in a useful way that we hope opens new research lines leading to an even brighter future for the field,” said Carlos Oscar Sánchez Sorzano, a group leader of the Biocomputing Unit of the National Center for Biotechnology / Instruct Image Processing Center in Madrid, Spain.

“The impact of this new algorithm goes well beyond the SARS-CoV-2 proteins and is widely applicable to any cryo-EM map below 3.5 angstroms, which are relatively common at present.” 

Researchers launched the DeepTracer website on July 21, 2020. The tool is fully automatic and freely available to anyone in the world, and allows users to upload their own 3DEM image or run data from an international collaboration called EMDataResource. This resource contains 3D images of macromolecular complexes, including COVID-19 and other viruses.

DeepTracer processes the image with an algorithm along with protein sequence information to predict where every atom is located. Compared to earlier state-of-the-art 3DEM modeling methods, results are more accurate, more complete, and quicker, arriving within a few minutes or hours depending on the image size.

“We can determine the atomic-level structure of a very large protein complex,” Si said. 

The team is working to make the tool more accurate and efficient to further contribute to COVID-19 vaccine development.

“Every day, I wake up in the morning and need to work on this project to make it faster, better and more accurate,” Si said. “I have a huge passion and dedication to it.” 

At Rose-Hulman Institute of Technology in Indiana, researchers are leveraging artificial intelligence to accelerate COVID-19 testing methods. A team is currently applying elements of AI, machine learning, and computational modeling to examine ways nanomaterials could improve the accuracy and response time of a promising saliva-based COVID-19 screening procedure.

If successful, the method would be safer and less costly than current methods and easily disposable. The device is currently still in development and clinical testing, but could provide COVID-19 test results in approximately 20 minutes. Researchers helped design the new testing approach using data-driven modeling techniques.

“Our team has joined at a very important time in this project,” said Assistant Professor Michael Jo. “We are working on constructing the first principle by simulation and to improve their pattern recognition of their test response.”

The method is currently doing real patient testing against SARS-CoV-2, and with further improvements, detecting a variety of influenza strains.

“The students' modeling and machine learning work, under the guidance of Dr. Jo and with our collaboration, will help us achieve a better understanding of the sensing characteristics within our project,” said Chief Technology Officer Ewa Kirkor.

“It will also help with the formulation of requirements, design, and building of the prototype for prompt, direct, and simultaneous detection and differentiation among multiple organisms.”

Researchers are using machine learning methods to increase the project’s limited sample size to improve detection accuracy. Additionally, the team is working to develop a framework that can reduce image data size while accelerating the speed of conventional machine learning and deep learning algorithms.

“Healthcare and medical diagnostics are among the major applications of machine learning. I hope my knowledge and work can help control this global pandemic,” said Xingheng Lin, a student in Jo’s lab.