Scientists at UW Bothell are using artificial intelligence and high tech mapping in the fight against coronavirus.
They’ve created a web-based software called Deeptracer it uses machine learning, artificial intelligence and high tech 3D imaging technology to predict the atomic structure of the coronavirus.
UW Bothell said the machine learning ‘analyzes a three dimensional image of a virus protein molecule and traces the connection of its atoms’.
“To fuse these technologies together and make the prediction very fast and accurate and more importantly automated,” said UW Bothell Assistant Professor of Computer Sciences and Principal Researcher Dong Si.
He said the proteins they are modeling and mapping are the red spikes that dot the iconic coronavirus sphere seen in artist renderings of the virus.
Si believes their research could prove to be i critical in the fight against the deadly virus and other future viruses, insisting if you can predict or model the atoms and molecules a virus is made up of, you can use it against it, particularly coronavirus.
“If we know all these detail structures, those pharmaceutical people and biologists and other people will be able to design a vaccine or drug,” said Si at his office on the UW Bothell campus today, “it’s the common sensing nature, that if you know the atom structure you know the function, so the 3D structure determines its functions, the root of problem of the pandemic is ,we don’t actually know the entire structure of the virus or all the related human cell molecular structures.”
The team launched a DeepTracer website and its free to use and scientist from around the world are using it and sharing data.
“That makes us feel that our work is important benefit to the people around the world,” said Si.
About 15 students most from UW Bothell, including one high school student have joined the research team – their work in part, includes: managing the website, tracking data, refining the teams’ algorithm and database engineering work.
Si said independent researchers put the accuracy of their modeling at 85 percent, but he insists that’s not good enough, he says the team is wants 100 percent.
