Humanity's quest for eternal youth has plagued us for millennia, and deep learning could bring it a step closer.
The mythological fountain of youth may one day become a reality, if tech titans such as Larry Ellison have their way. The Oracle co-founder has ploughed $430 million into life-extending research, putting much of his hope on research into the hormone DHEA, which some scientists believe slows down the ageing process.
Dr Alex Zhavoronkov believes the secret could be found in deep learning.
The Latvia-born anti-ageing expert studied computer science and built a career in IT before he swapped it for one in biotech.
In 2014, he founded Insilico Medicine, a Baltimore-based company that specialises in using artificial intelligence for drug discovery and ageing research.
From their base at John Hopkins University, Zhavoronkov and his colleagues developed algorithms that analyse the minute differences between normal and diseased human tissue.
"We started using our pathway algorithms to feed the data into neural networks and started building pretty accurate predictors and classifiers of cancer, Alzheimer's, Parkinson's, and also ageing," says Zhavoronkov.
By using deep learning to aggregate data from genetic tests, radiology images, blood and urine tests and photographs, Insilico found a way to predict human age.
"When you train neural networks on age, you can later re-train them on the various diseases when you have a lot of samples available. We started extracting the most important features that play a role in various diseases and in ageing and starting developing molecules for those features."
Insilico recently partnered with GlaxoSmithKline to explore how these molecules can aid the drug discovery process by targeting specific aspects of diseases.
The healthcare data marketplace
Among the technology developed by Silico is a platform called Young.AI, which integrates the multiple predictors of a user’s age to help them manage their health and lifestyle over time.
Its search for more sources of information that can predict ageing has led Insilico to the type of privacy issues shared by many data-driven companies.
"A lot of people don't regard pictures as a very valuable data type and they post them on Facebook and Google freely," says Zhavoronkov.
"They post their pictures everywhere but they are very concerned about their genomes, for example. People don't like to share their genomes."
In the case of identifying ageing, a picture can be worth more than a genome. Deep neural networks in conjunction with clinical geneticists can evaluate a picture to understand the person's age, their sex, social status and more than 1,000 genetic disorders, such as Down's syndrome and Hutchinson-Gilford progeria syndrome.
If these images cover an individual’s life over decades and can be assessed in relation with medical information such as blood tests they can combine to form a very valuable dataset. But for deep learning predictions to accurately assess a vast array of diseases in a diverse global population, it will need to be trained on an enormous dataset.
Insilico believes it can build this by creating a digital data marketplace, where users can upload and sell their health data.
The company has partnered with Bitfury to use its Exonum blockchain platform to store and secure this information.
Researchers at the company published a paper in a peer-reviewed journal titled "Converging blockchain and next-generation artificial intelligence technologies to decentralize and accelerate biomedical research and healthcare", which they hope will attract the support of more academics.
"I think that it's the only way to return the control over human life data to people,” says Zhavoronkov. "Currently, there is a major disconnect between consumers, regulators and companies.
"If they want to make their data available, they can. If they don't want to make the data available, they should be able to block other people from seeing it. I think the introduction of this marketplace will put more discipline into what is happening right now, because right now, it's the wild west, essentially."
He imagines biotech companies using it to develop healthcare treatments by training algorithms on the data, and consumer companies using it to analyse understand customer lifestyles by analysing their photographs to make recommendations on how they could be more healthy.
"Every company, every government should be striving and craving to make people healthier and more productive, because ageing is one of the major threats to the global economy."
It's also an increasingly lucrative market. Estimates by market research firm Frost & Sullivan suggest the global revenues for healthcare AI could hit $6.7 billion by 2021.
"We naturally assume that we need, from time to time, to get sick and at some point in time, die of ageing," says Zhavoronkov. "But with the advances in AI and also with the availability of more data, we have all the tools to cure all diseases and possibly within our lifetime."