To create a new drug today, scientists have to test tens of thousands of components to understand how they interact. And that’s not the hard part. After a substance found effective against the disease, he will undergo three different phases of clinical trials and to obtain regulatory approval.
According to estimates, on average, one new drug came on the market, you need 1000 people, 12-15 years and about $ 1.6 billion. It seems to be a better way — and he is thought to have appeared. Last week, scientists published a paper which describes in detail the artificial intelligence system created to assist in the search for new drugs. It needs significantly less amount of time and money spent in the process.
The system is called AtomNet and made it a startup from San Francisco called AtomWise. The technology is designed to streamline the initial stage of drug discovery, which involves the interaction of various molecules with each other, in particular, scientists need to determine what molecules will bind and how strongly. They use a method of trial and error, going through tens of thousands of components, both natural and synthetic.
AtomNet shorten this process by using techniques of deep learning to predict how molecules and how likely they are to form a bond. Software study of molecular interaction, recognizing patterns, like the AI learns to recognize images.
Remember the three-dimensional model of atoms, which many do in high school out of foam and tubes to represent the relationships between protons, neutrons, and electrons? AtomNet uses a similar three-dimensional models of molecules, including information about their structure, to predict their biological activity.
Says chief operating officer AtomWise Alexander Levi, “you can take the interaction between drug and biological system and to decompose it into smaller interactive groups. If you study enough historical examples of molecules that can be pretty fast to make accurate predictions”.
“Quickly” might even be an understatement. Reportedly, AtomNet can go through a million compounds per day. Applying modern methods, this would take months.
AtomNet can’t invent a new drug or even say for sure whether combination of two molecules effective medication. But it can predict how likely a certain compound works against a specific disease. Scientists use these predictions to narrow down thousands of options to tens, to focus testing where positive results are most likely.
This software has already proven itself, helping to create new drugs to treat Ebola and multiple sclerosis. The last drug was licensed to a British pharmaceutical company and the drug against Ebola was presented in a peer-reviewed journal for further analysis.
Although AtomNet is a promising technology that will accelerate the discovery of new drugs, it is worth noting that the future of medicine is also moving towards a proactive rather than a reactive approach; instead of trying to invent drugs just for the treatment of sick people, attention shifted to the careful monitoring of health status and take necessary steps that will not let us sick in the first place.
Last year, the Fund’s Zuckerberg gave $ 3 billion to search for the “cure for all diseases”. This is an ambitious and somewhat quixotic goal that nevertheless deserves respect. In another example of the movement toward proactive health XPRIZE Foundation recently awarded $ 2.5 million device designed to help diagnose at home and personal health monitoring. Proactive technology in health care is likely to be to develop and grow in popularity.
But this does not mean that reactive medical care will remain in place. In fifty or a hundred years people will still get sick and need medicine that will cure them. AtomNet — a first of its kind software. But very soon there will be other methods of applying artificial intelligence on this path.