{"id":153,"date":"2019-02-07T15:53:59","date_gmt":"2019-02-07T18:53:59","guid":{"rendered":"http:\/\/ethosclinical.com.br\/?p=153"},"modified":"2019-02-07T15:56:39","modified_gmt":"2019-02-07T18:56:39","slug":"153-2","status":"publish","type":"post","link":"http:\/\/ethosclinical.com.br\/en\/153-2\/","title":{"rendered":"Making New Drugs With a Dose of Artificial Intelligence"},"content":{"rendered":"\n

https:\/\/www.nytimes.com\/2019\/02\/05\/technology\/artificial-intelligence-drug-research-deepmind.html?partner=rss&emc=rss<\/a><\/p>\n\n\n\n

Making\nNew Drugs With a Dose of Artificial Intelligence<\/h1>\n\n\n\n

Feb.\n5, 2019<\/p>\n\n\n\n

\"\"
T.M.\nDetwiler\n\n<\/figcaption><\/figure>\n\n\n\n

SAN\nFRANCISCO \u2014 You can think of it as a World Cup of biochemical research.\n\n<\/p>\n\n\n\n

Every two years, hundreds of scientists\nenter a global competition. Tackling a biological puzzle they call \u201cthe protein\nfolding problem,\u201d they try to predict the three-dimensional shape of proteins\nin the human body. No one knows how to solve the problem. Even the winners only\nchip away at it. But a solution could streamline the way scientists create new\nmedicines and fight disease.<\/p>\n\n\n\n

Mohammed AlQuraishi, a biologist who\nhas dedicated his career to this kind of research, flew in early December to\nCancun, Mexico, where academics were gathering to discuss the results of the\nlatest contest. As he checked into his hotel, a five-star resort on the\nCaribbean, he was consumed by melancholy.<\/p>\n\n\n\n

The contest, the Critical Assessment of\nStructure Prediction, was not won by academics. It was won by DeepMind, the\nartificial intelligence lab owned by Google\u2019s parent company.<\/p>\n\n\n\n

\u201cI was surprised and deflated,\u201d said\nDr. AlQuraishi, a researcher at Harvard Medical School. \u201cThey were way out in\nfront of everyone else.\u201d<\/p>\n\n\n\n

DeepMind specializes in \u201cdeep\nlearning,\u201d a type of artificial intelligence that is rapidly changing drug\ndiscovery science. A growing number of companies are applying similar methods\nto other parts of the long, enormously complex process that produces new\nmedicines. These A.I. techniques can speed up many aspects of drug discovery\nand, in some cases, perform tasks typically handled by scientists.<\/p>\n\n\n\n

\u201cIt is not that machines are going to\nreplace chemists,\u201d said Derek Lowe, a longtime drug discovery researcher and\nthe author of In the Pipeline, a widely read blog dedicated to drug discovery.\n\u201cIt\u2019s that the chemists who use machines will replace those that don\u2019t.\u201d<\/p>\n\n\n\n

After the conference in Cancun, Dr.\nAlQuraishi described his experience in a blog\npost<\/a>. The melancholy he felt after losing to DeepMind gave way to\nwhat he called \u201ca more rational assessment of the value of scientific\nprogress.\u201d<\/p>\n\n\n\n

But he strongly criticized big\npharmaceutical companies like Merck and Novartis, as well as his academic\ncommunity, for not keeping pace.<\/p>\n\n\n\n

\u201cThe smartest and most ambitious\nresearchers wanting to work on protein structure will look to DeepMind for\nopportunities instead of Merck or Novartis,\u201d he wrote. \u201cThis fact should send\nchills down the spines of pharma executives, but it won\u2019t, because they\u2019re\nclueless, rudderless, and asleep at the helm.\u201d<\/p>\n\n\n\n

The big pharma companies see the\nsituation differently. Though Merck is not exploring protein folding because\nits researchers believe its potential impact would be years away, it is\napplying deep learning to other aspects of its drug discovery process.<\/p>\n\n\n\n

\u201cWe have to connect so many other\ndots,\u201d said Juan Alvarez, associate vice president of computational and\nstructural chemistry at Merck.<\/p>\n\n\n\n

In the spring of 2016, after making\nheadlines with A.I. systems that played complex games like the ancient board\ngame Go, DeepMind researchers were looking for new challenges. So they held a\n\u201chackathon\u201d at company headquarters in London.<\/p>\n\n\n\n

Working with two other computer\nscientists, the DeepMind researcher Rich Evans homed in on protein folding.\nThey found a game that simulated this scientific task. They built a system that\nlearned to play the game on its own, and the results were promising enough for\nDeepMind to greenlight a full-time research project.<\/p>\n\n\n\n

The protein folding problem asks a\nstraightforward question: Can you predict the physical structure of a protein \u2014\nits shape in three dimensions?<\/p>\n\n\n\n

If scientists can predict a protein\u2019s\nshape, they can better determine how other molecules will \u201cbind\u201d to it \u2014 attach\nto it, physically \u2014 and that is one way drugs are developed. A drug binds to\nparticular proteins in your body and changes their behavior.<\/p>\n\n\n\n

In the latest contest, DeepMind made\nthese predictions using \u201cneural networks,\u201d complex mathematical systems that\ncan learn tasks by analyzing vast amounts of data. By analyzing thousands of\nproteins, a neural network can learn to predict the shape of others.<\/p>\n\n\n\n

This is the same deep learning\ntechnology that recognizes faces in the photos you post to Facebook. Over the\npast decade, the technology has reinvented a wide range of internet services, consumer products<\/a>,\nrobotic devices<\/a> and\nother areas of scientific research<\/a>.<\/p>\n\n\n\n

Many of the academics who competed used\nmethods that were similar to what DeepMind was doing. But DeepMind won the\ncompetition by a sizable margin \u2014 it improved the prediction accuracy nearly\ntwice as much as experts expected from the contest winner.<\/p>\n\n\n\n

DeepMind\u2019s victory showed how the\nfuture of biochemical research will increasingly be driven by machines and the\npeople who oversee those machines.<\/p>\n\n\n\n

This kind of A.I. research benefits\nfrom enormous amounts of computing power, and DeepMind can lean on the massive\ncomputer data centers that underpin Google. The lab also employs many of the\nworld\u2019s top A.I. researchers, who know how to get the most out of this\nhardware.<\/p>\n\n\n\n

\u201cIt allows us to be much more creative,\nto try many more ideas, often in parallel,\u201d said Demis Hassabis, the chief\nexecutive and a co-founder of DeepMind, which Google acquired for a reported\n$650 million in 2014.<\/p>\n\n\n\n

Universities and big pharmaceutical\ncompanies are unlikely to match those resources. But thanks to cloud computing\nservices offered by Google and other tech giants, the price of computing power\ncontinues to drop. Dr. AlQuraishi urged the life-sciences community to shift\nmore attention toward the kind of A.I. work practiced by DeepMind.<\/p>\n\n\n\n

Some researchers are already moving in\nthat direction. Many start-ups, like Atomwise in San Francisco and Recursion in\nSalt Lake City, are using the same artificial intelligence techniques to\naccelerate other aspects of drug discovery. Recursion, for instance, uses\nneural networks and other methods to analyze images of cells and learn how new\ndrugs affect these cells.<\/p>\n\n\n\n

The big pharma companies are also\nbeginning to explore these methods, sometimes in partnership with start-ups.<\/p>\n\n\n\n

\u201cEveryone is trending up in this area,\u201d\nsaid Jeremy Jenkins, the head of data science for chemical biology and\ntherapeutics at Novartis. \u201cIt is like turning a big ship, and I think these\nmethods will eventually scale to the size of our entire company.\u201d<\/p>\n\n\n\n

Mr. Hassabis said DeepMind was\ncommitted to solving the protein folding problem. But many experts said that\neven if it was solved, more work was needed before doctors and patients\nbenefited in any practical way.<\/p>\n\n\n\n

\u201cThis is a first step,\u201d said David Baker,\nthe director of the Institute for Protein Design at the University of\nWashington. \u201cThere are so many other steps still to go.\u201d<\/p>\n\n\n\n

As they work to better understand the\nproteins in the body, for instance, scientists must also create new proteins\nthat can serve as drug candidates. Dr. Baker now believes that creating\nproteins is more important to drug discovery than the \u201cfolding\u201d methods being\nexplored, and this task, he said, is not as well suited to DeepMind-style A.I.<\/p>\n\n\n\n

DeepMind researchers focus on games and\ncontests because they can show a clear improvement in artificial intelligence.\nBut it is not clear how that approach translates to many tasks.<\/p>\n\n\n\n

\u201cBecause of the complexity of drug\ndiscovery, we need a wide variety of tools,\u201d Dr. Alvarez said. \u201cThere is no\none-size-fits-all answer.\u201d<\/p>\n","protected":false},"excerpt":{"rendered":"

https:\/\/www.nytimes.com\/2019\/02\/05\/technology\/artificial-intelligence-drug-research-deepmind.html?partner=rss&emc=rss Making New Drugs With a Dose of Artificial Intelligence Feb. 5, 2019 SAN FRANCISCO \u2014 You can think of it as a World Cup of biochemical research. Every two years, hundreds of scientists enter a global competition. Tackling a biological puzzle they call \u201cthe protein folding problem,\u201d they try to predict the three-dimensional shape […]<\/p>\n","protected":false},"author":2,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":true,"template":"","format":"standard","meta":{"spay_email":"","jetpack_publicize_message":""},"categories":[1],"tags":[],"jetpack_featured_media_url":"","jetpack_publicize_connections":[],"jetpack_shortlink":"https:\/\/wp.me\/p73690-2t","_links":{"self":[{"href":"http:\/\/ethosclinical.com.br\/en\/wp-json\/wp\/v2\/posts\/153"}],"collection":[{"href":"http:\/\/ethosclinical.com.br\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"http:\/\/ethosclinical.com.br\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"http:\/\/ethosclinical.com.br\/en\/wp-json\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"http:\/\/ethosclinical.com.br\/en\/wp-json\/wp\/v2\/comments?post=153"}],"version-history":[{"count":2,"href":"http:\/\/ethosclinical.com.br\/en\/wp-json\/wp\/v2\/posts\/153\/revisions"}],"predecessor-version":[{"id":155,"href":"http:\/\/ethosclinical.com.br\/en\/wp-json\/wp\/v2\/posts\/153\/revisions\/155"}],"wp:attachment":[{"href":"http:\/\/ethosclinical.com.br\/en\/wp-json\/wp\/v2\/media?parent=153"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"http:\/\/ethosclinical.com.br\/en\/wp-json\/wp\/v2\/categories?post=153"},{"taxonomy":"post_tag","embeddable":true,"href":"http:\/\/ethosclinical.com.br\/en\/wp-json\/wp\/v2\/tags?post=153"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}