Personalized healthcare is as heavily hyped as AI itself, and so it is hardly surprising that the combination of the two, has attracted a host of start-ups. The ambition is both diagnostic and prescriptive, aiming to identify not just current medical conditions but ones that the patient might succumb to in future on the basis of lifestyle and genetic factors. Treatments would also be determined not just by genetics of the patient but also molecular details of the disease.
This is a big data problem of immense proportions and AI figures strongly in the attempt to identify characteristic signatures of diseases or risks, as well as matching these to therapies. Of all the start-ups attacking this field from various directions, none has made a greater impact in a short time than China’s iCarbonX, already a unicorn defined as being young, privately held and having a valuation over $1 billion, just over two years after its foundation in October 2015. We might ponder that there is no agreed definition of start-up itself but a common consensus might be a privately held company less than five years old, so iCarbonX still qualifies, while say Uber approaching its 10th birthday no longer does.
iCarbonX has several advantages over some other start-ups in personalized healthcare, beyond the $600 million it has raised so far in finance. Firstly, it is Chinese, which in healthcare especially gives it access to data on an unprecedented scale, not just because of the country’s 1.4 billion population, but also because a lot of that data would be denied for privacy or regulatory reasons in the US and Europe. China’s government is friendly, in that regard.
Data is what drives success in AI generally and especially healthcare, where the number of variables and sensitivity to multiple conditions requires massive statistical power to make clear deductions in many cases. Secondly, the company is heavily rooted in healthcare, with some leading practitioners on board – above all its founder and CEO Jun Wang, who previously co-founded the influential the Beijing Genomics Institute (BGI). In 2010, when Wang was still at the helm, Nature magazine reported that BGI’s human genome sequencing output could in theory surpass that of all genomics company and other initiatives in the United States combined, pointing again to the potential power of Chinese healthcare data.
A third highly relevant factor is that Wang realized from the outset that success would also depend on assembling as wide a collation as possible, drawing on expertise in the US and Europe through collaborations. Wang was well aware that while China might have a suitably relaxed regulatory regime, it lacked the relatively organized digital patient health records found in the US and some European countries. As a result, the company, based in China’s silicon corridor in Shenzhen on the mainland adjacent to Hong Kong, reached out to various US-based start-ups specializing in health data and personalized medicine startups. They in turn were enticed by the opportunity to cut loose on China’s potentially massive if still disorganized healthcare data. This included the likes of SomaLogic and HealthTell, which are launching their genetic tests commercially in China.
SomaLogic from Boulder, Colorado, is a key partner, that has developed a chip that can identify and track the activity of around 4,200 proteins simultaneously. This is significant progress, noting that in June 2016 a commercially available precursor to this chip reading 1,130 proteins achieved a milestone by predicting which subset of heart-attack patients would have a recurrence. This involved measuring the activity of nine blood proteins in more detail.
The firm’s scientific ambition is to be the leader in predictive medicine and especially take the ability to analyze genomics and personalized protein data to a new level. Wang, in an echo of other healthcare start-ups, explained that he had been motivated to start iCarbonX out of frustration with existing predictive capabilities, especially relating to analysis of variations between the genes of individuals. Apart from a number of rare disease associated with single genes, such as cystic fibrosis of Huntingdon’s Syndrome, prediction tended to be confined to statements such as a patient having a 5% greater chance than average of contracting say heart disease or a particular cancer by a given age. Such information was of limited value in personalized healthcare.
It remains to be seen how quickly more precise prediction resulting from analysis of variations (known as alleles) across multiple genes can be made, or whether iCarbonX will achieve the leadership role Wang craves. It certainly has the data, but questions remain over whether complex conditions including many cancers and metabolic disorders like type 2 diabetes and Crohn’s Disease will ever be highly predictable from genetic data. These conditions may be too sensitive to external environmental conditions, such that it will not be sufficient just to squeeze the noise out of the data through massive statistical scale.
It is clear though the company is spreading its bets and developing not just one platform but an ecosystem of products with no rush to commercialize, given the backing it has. It is therefore closer to the Amazon than Google model. This does mean it is competing on a number of fronts, bringing it up against the big US tech firms such as IBM, Microsoft, Apple as well as Google and Amazon themselves for which AI-based healthcare is a major target. There are also numerous smaller firms or start-ups such as Arivale in Seattle working on similar technology.
The main risk is not so much from competitors but the possible failure of precision medicine to live up to its promise, at least in the immediate future. Even iCarbonX does not have unlimited patience or resources. It is a good bet though to become a major global AI healthcare player given its pedigree, roots and alliances.