Alex Zhavoronkov, PhD, a Board Member and Editor of the Journal Aging, wants you to live as long as possible and be younger at every level. “One hundred and fifty is just a number which I recommend everyone use just for subconscious life planning,” said Zhavoronkov. “The longer you plan to live, the longer you will live and the younger you will behave.” Zhavoronkov believes there is a psychological component to aging, and longer lifespan horizons are something we can reasonably achieve, but we first have to believe we can.
Zhavoronkov is the Founder, CEO and CSO of Insilico Medicine, a company dedicated to providing artificial intelligence (AI) for drug discovery and aging research, and is a pioneer working in biotechnology, regenerative medicine, and aging economics. Zhavoronkov and his team provide services to academia, pharmaceutical and cosmetics companies such as advanced deep learning solutions and custom drug discovery and biomarker engines. In other words, Zhavoronkov is actively researching how drug interactions affect lifespan in people with differing health profiles in the hope of improving health into old age and longevity overall.
Creating deep neural networks to predict health interactions on chronological age
By creating AI models, Zhavoronkov employs deep learning and deep neural network concepts that have been around since the 1980s. Essentially, deep neural networks are networks of neurons that are represented by computer algorithms that change in response to unstructured data. A model is created in a computer that is patterned after the human brain and human system. Deep neural networks are created from the resulting algorithm, and model the response to stimuli the way a living human would. Researchers are able to use these AI models to test the effects of drugs on lifespan without putting an actual person’s life at risk.
“Deep learning, or deep neural networks, have gained prominence in the last five years, achieving super human performance text recognition, and voice recognition, with no human intervention,” said Zhavoronkov. “These networks learn tasks and pattern recognition, and we pioneered their use in aging research, training deep neural networks to predict human chronological age.”
Using blood samples to applying deep learning to aging
Zhavoronkov’s team performs deep learning techniques in two ways. Top down, he is able to understand the aging process better and formulate predictors for novel targets that might be addressable by drugs. Bottom up, he designs normal molecules using deep learning.
“We take a very large number of blood tests, and annotated for age, we can build comprehensive and accurate predictors of chronological age using this simple beta type,” said Zhavoronkov.
These simple clinical blood tests are used to create deep neural networks that capture the most important relevant neural markers. This enables Zhavoronkov to extract the most relevant features for predicting aging in sample groups, and understand what drives causality in aging.
“It is serendipitous that aging has been my area of interest for a long time,” said Zhavoronkov.
‘In that the recent technological advances in AI have also advanced our knowledge of aging.”
Everyone ages, making aging the most universal feature for multi model integration
Zhavoronkov is a leader in biomarker discovery, and AI has given him the ability to come up with novel techniques for looking at aging from multiple perspectives.
“Not every person has cancer, diabetes, or Parkinson’s, for example,” said Zhavoronkov. “But all patients have age. Every living being on this planet has age. Aging as a disease manifests itself as deconvolution, and with AI, we can train deep neural networks to understand the most relevant features important to aging.”
We age on many levels: molecular, cellular, tissue, organs, and psychological. We also change on many levels so researchers need to work with many beta types to understand aging.
“If you just study cancer or diabetes, it is difficult to find people who are similar and share similar diseases,” said Zhavoronkov. “When you are working with aging, everyone has age. We use this single feature for multi model data integration, and integrating transcriptome (all the messenger RNA molecules expressed from the genes of an organism) with data from a wearable device. But in cancer, those data types would not be integrable, and instead it might be possible to use wearable devices to detect cancer.
“But you can predict age using activity patterns, and it is possible to predict your age from a blood test,” said Zhavoronkov. “By using age as universal feature in multi model integration, we are able to create virtual humans that change in time on many levels.”
To illustrate how AI can predict aging, in one experiment Zhavoronkov used a family of monkeys and asked people to guess their age. The task was relatively easy, and most people were able to guess the age of the monkeys even though they had never seen a large number of monkeys. The reason they could accurately guess is because they have seen humans and they extrapolated the features of humans on to monkeys.
“We do this for may data types, and we use age to predict disease risk,” said Zhavoronkov. “For example, we can get a good picture of the risk and try to build a pathway based on our understanding of the features relevant to Alzheimer’s.”
Mimicking drug effects with nutraceuticals
Zhavoronkov’s recent study in Aging describes his effort to identify nutraceuticals, which are safe, naturally occurring compounds, usually derived from food, that offer extra health benefits above the basic nutrition in food. Nutraceuticals can be used to mimic the the anti‐aging effects of metformin and rapamycin without adverse effects. Metformin and rapamycin are two FDA‐approved mTOR inhibitors that have shown to reduce the burden of chronic disease and extend human healthspan, as well as exhibiting significant anti‐cancer and anti‐aging properties beyond their current clinical applications. But due to limits for off-label use approval for those drugs, Zhavoronkov has turned to nutraceuticals that do not require FDA approval.
It has been difficult to get people who take nutraceuticals to report their states of health back to the companies that manufacture them. So Zhavoronkov partnered with a company called Life Extension to compound several molecules together to create and start selling a nutraceutical.
“We partnered with Life Extension because they provide their customers with a voluntary blood test in order to get data back to see how the nutraceuticals are performing,” said Zhavoronkov. The data is analyzed to look for any expected or unexpected effects of nutraceuticals on real humans in terms of anti-aging effects. Zhavoronkov hopes to innovate nutraceuticals-based clinical trials.
To get data, Zhavoronkov is also beta testing a platform called Young.ai where people can voluntarily upload their blood tests and photos to create good AI predictors prior to clinical trials. He currently has a pipeline of molecules for cancer and fibrosis, and is developing molecules for dermatological indications.
“You need to see it to believe it,” said Zhavoronkov. “Scientists are developing AI and deep learning systems so advanced that they allow you to see with your eyes if your algorithm is working.”
Inspired by out of the box thinking
When asked what inspires and drives the work he does on aging, and why he values his role as an Editor of Aging, Zhavoronkov said “I enjoy working with really brilliant people, people who work at the intersection of AI and biology and who think outside the box. There are very few of those people in the world, and at Insilico we managed to bring in people who are at the top of the field.
“I am working with a new generation of biogerontology focused on making a difference, not only in publishing papers, but also focusing on changing the way we treat and research aging,” said Zhavoronkov.
“It’s a great pleasure to see that research in aging is advancing quickly, as is AI,” added Zhavoronkov. “This industry is changing quickly and people are proposing and implementing new methods every day. In the next five years we will see a major transformation in how we look at aging.”