Chinese Health App Arrives Access to a Large Population Used to Sharing Data Could Give Icarbonx an Edge Over Rivals

Home , Fan Hui, Gu Li (Go player), Master (software), Tygem

NEWS IN FOCUS ASTROPHYSICS Legendary CHEMISTRY Deceptive spice POLITICS Scientists spy ECOLOGY New Zealand Arecibo telescope faces molecule offers cautionary chance to green UK plans to kill off all uncertain future p.143 tale p.144 after Brexit p.145 invasive predators p.148 ICARBONX

Jun Wang, founder of digital biotechnology firm iCarbonX, showcases the Meum app that will use reams of health data to provide customized medical advice.

BIOTECHNOLOGY Chinese health app arrives Access to a large population used to sharing data could give iCarbonX an edge over rivals.

BY DAVID CYRANOSKI, SHENZHEN medical advice directly to consumers through another $400 million had been invested in the an app. alliance members, but he declined to name the ne of China’s most intriguing biotech- The announcement was a long-anticipated source. Wang also demonstrated the smart- nology companies has fleshed out an debut for iCarbonX, which Wang founded phone app, called Meum after the Latin for earlier quixotic promise to use artificial in October 2015 shortly after he left his lead- ‘my’, that customers would use to enter data Ointelligence (AI) to revolutionize health care. ership position at China’s genomics pow- and receive advice. The Shenzhen firm iCarbonX has formed erhouse, BGI, also in Shenzhen. The firm As well as Google, IBM and various smaller an ambitious alliance with seven technology has now raised more than US$600 million in companies, such as Arivale of Seattle, Wash- companies from around the world that special- investment — this contrasts with the tens of ington, are working on similar technology. But ize in gathering different types of health-care millions that most of its rivals are thought Wang says that the iCarbonX alliance will be data, said the company’s founder, Jun Wang, to have invested (although several big play- able to collect data more cheaply and quickly. on 5 January at the Digital Life Summit, which ers, such as Google, have not disclosed exact The tools offered by alliance partners are was hosted by iCarbonX. figures). Initial funding of some $200 million cutting edge and the set-up will allow different The alliance will use algorithms to analyse came mostly from Shenzhen-based Tencent, types of data to be integrated seamlessly; more- reams of genomic, physiological and behav- which owns the social-media application over, China’s large population is already used to ioural data and provide customized health and WeChat. At the summit, Wang said that roughly sharing information through WeChat and

NEWS IN FOCUS

other social media, Wang notes. “No one uses microarrays of some 330,000 protein that should reduce noise in the data and allow will be able to collect at the same scale that I am fragments to fish antibodies from a blood patterns to emerge, says Bernard Munos, a sen- doing,” he says. Wang is confident that he can sample to answer questions about disease ior fellow at FasterCures, a drug-development get samples and data from one million people progress, allergies and vaccine effectiveness advocacy organization in Washington DC. in five years, which in turn will lead to a more (J. B. Legutki et al. Nature Commun. 5, 4785; “This is very important given the number of informed AI. 2014). And another, PatientsLikeMe of Cam- variables they will be dealing with.” His plans are driven by frustration with bridge, Massachusetts, asks its 500,000 or so But he worries about the chaotic behav- genomics. The largest genomic studies offer users to upload to a website less-clear-cut data iour of some of the systems. Alliance member only subtle hints about an individual’s suscep- about pain levels, sleep and fatigue. These are General Automation Lab Technologies of San tibility to disease — such as pinpointing a gene combined with medical data, behaviour pat- Francisco, California, will provide personal- that makes an individual only one or two per terns and the users’ experiences of diseases and ized assessments of how microbes living on cent more likely to develop heart disease. So, in drugs to find patterns someone’s body may affect their health. But addition to mining its customer’s genomes, the “He sucks you in that predict such the microbiome is poorly understood and in iCarbonX alliance will scour biological mol- with a vision of ‘immeasurables’. a constant state of flux, says Munos. “There is ecules from various tissues to provide a more what can be, and The end result will a lot of randomness in biology at the individ- accurate and actionable picture of someone’s then he makes it be an unwieldy set ual level that will be hard to capture, let alone health. Wang chose alliance members for their happen.” of data from vari- model. I wish them well, but, at the moment, I promise in mining such signals. ous sources, which is am guarded as to their chance of success.” SomaLogic of Boulder, Colorado, for exam- why Wang and a team at iCarbonX are devel- Others have faith in Wang. “There’s no ple, has a chip that can measure some 4,200 pro- oping algorithms to understand how these bullshit. He sucks you in with a vision of what teins simultaneously. In June, researchers using variables correlate with healthy or diseased can be, and then he makes it happen,” says a commercially available precursor that reads states. The Meum app enables users to enter Larry Gold, founder and chairman of alliance 1,130 proteins were able to predict which their meals and activity levels, as well as any member SomaLogic. subset of heart-attack patients would have a physiological or vital-sign data, and gives Wang is keenly aware that the success of the recurrence by measuring the activity of nine advice on what to eat, when to sleep and how venture will depend on its users’ readiness to blood proteins (P. Ganz et al. J. Am. Med. Assoc. active they should be. submit data and heed the advice the app gives. 315, 2532–2541; 2016). Views are mixed about how well the venture He said during one of his many toasts during Another alliance member, HealthTell of will work. The alliance will provide iCarbonX the summit: “Meum might tell me not to drink, San Ramon, California, makes a chip that with a collection of high-quality indicators but I don’t have to listen.” ■

ARTIFICIAL INTELLIGENCE Google secretly tested AI bot Updated version of Google DeepMind’s AlphaGo program revealed as mystery online player.

BY ELIZABETH GIBNEY platforms. And on 4 January, Google Deep- master the game, something that both Chinese Mind chief executive Demis Hassabis revealed and South Korean scientists have said they are mystery player causing a stir in the on Twitter that Master(P) is a new prototype attempting to do. world of the complex strategy game version of AlphaGo. The “unofficial” games Hassabis said that the new version of Go has been revealed as an updated were designed to test the prototype, he said: AlphaGo would play official, full-length games Aversion of AlphaGo, the artificial-intelligence “We’re excited by the results and also by what later this year. How strong it will be in more (AI) program created by Google’s London- we and the Go community can learn from high-profile tournaments remains unclear, based AI firm, DeepMind. some of the innovative and successful moves because the rules of such matches differ from Known only by the name Master(P), the played by the new version of AlphaGo.” those of matches played in online forums. anonymous player has beaten the world’s best Playing on the online servers Tygem and Online games are usually played at a faster at Go in a string of online games since late FoxGo, Master(P) played more than 50 games, pace, which favours the computer over December, including defeating current world winning them all — except perhaps for one humans, says Rémi Coulom, a freelance number one, 19-year-old Ke Jie. game, which, according to some reports, was developer of Go programs based in Lille, Go is regarded as the most complex board deemed a tie only because the network connec- France. “But still, I expect a strong correlation game ever invented, and is famously difficult tion of the opponent, the Go professional Chen with performance in serious slow tournament for computers to crack. But last year, AlphaGo Yaoye, timed out. “It’s extremely impressive games,” he adds. showcased the strength of AI software when whoever/whatever it is,” said British Go player AlphaGo has played only around a dozen it stunned the Go world, first by defeating a Jon Diamond ahead of the announcement. public games, so Google DeepMind’s decision professional human player, Fan Hui, and then After losing to Master(P), Chinese profes- to trial its latest version in the open will allow by going on to beat one of the Go world’s top sional Gu Li offered a reward of 100,000 yuan Go players to study more of its moves. players, Lee Sedol. (US$14,400) to any human who could beat the “I personally think it’s fantastic that there are Fellow players had a hunch that Master(P) mysterious player. all these games for people to look at and study. was probably also an AI program. It came Although AlphaGo was rumoured to be There are lots of moves that are really new and out of nowhere to win dozens of consecutive behind the bot, many observers also suspected surprising,” says Niall Cardin, a UK-based Go quick-fire games across two separate online that another team had created an AI that could player. ■

142 | NATURE | VOL 541 | 12 JANUARY©201 72017Mac millan Publishers Li mited, part of Spri nger Nature. All ri ghts reserved.