AtomNet, a system created by San Francisco-based startup Atomwise, is designed to help with the goal of curing major diseases by predicting the bioactivity of small molecules using a deep learning neural network. The result? New drugs, invented by robots.
“AtomNet is an artificial intelligence system that we use to help design and discover new compounds for medical research,” Dr. Kong Nguyen, Atomwise’s senior scientist, told Digital Trends. “It works by analyzing the biological structures and processes involved with diseases like cancer, multiple sclerosis, and Ebola, and simulating how potential medicines will interact with them. Research labs at universities and pharmaceutical companies [can then] take AtomNet predictions, synthesize them in the real world, and test them to discover their medical value.”
A quick glance at any medical history textbook will reveal that humans haven’t done too badly when it comes to drug discovery. What makes AtomNet exciting, however, is its ability to not just learn from millions of example of past data about unsolved diseases, but to do this incredibly quickly.
How quickly?
“AtomNet performs its work extremely fast, screening about 1 million compounds each day,” Nguyen continued. “That speed, combined with its high accuracy, allows us to do new and interesting kinds of research. For example, the AIMS program for academics uses AtomNet to allow any academic researcher to consider 10 million compounds for their diseases. This would be extremely hard them to do by other means, potentially taking millions of dollars and many years of physical experimentation.”
AtomNet has already created promising drugs for battling multiple sclerosis and Ebola. One of these has been licensed to a pharmaceutical company in the U.K., while the Ebola drug is set to be submitted to a peer-reviewed journal for further appraisal.
As Nguyen points out, though, it may still be a little while before we’re using drugs designed by bots.
“The discoveries Atomwise helps make will certainly take some time to find their way through clinical trials and eventually regulatory approval,” he said. “Today, the total process for a single new medicine takes approximately 15 years, on average.”
But the hope is that AI can also help speed this up. “We are optimistic that Atomwise can help shorten that considerably, helping to reverse the trend in recent decades towards longer timelines in the discovery and development of new treatments,” he noted.
We’ll keep our fingers crossed!
Editors' Recommendations
- Revisiting the rise of A.I.: How far has artificial intelligence come since 2010?
- Designing new drugs takes years, but A.I. could help reduce that to days
