How is AI decoding aging?

In the neoliberal era, where productivity and constant self-optimization seem to be the norm to enable an individuals’ competitiveness on the market, people look in various ways for self-care. Wearables, mobile apps, and other sensors fueling the quantified self-movement, give us hope that mindfulness of our steps, sleep and diet will help us live better and longer. Biohackers go a step further and hack natural mechanisms of the human body. As youth and eternal life have been glorified throughout human history, these interventions are attractive. Increasing personalization of health and wellness data analysis and data-driven recommendations are further boosting optimism and hope of a more vital old age. But to what extent can artificial intelligence (AI) solve aging?

First of all: what is next generation AI?

Artificial intelligence can be classified in different techniques: machine learning, deep learning, reinforcement learning, general adversarial networks, transfer learning and meta-learning (so-called next-generation AI).

What are these techniques again?

Machine learning refers to algorithms that can learn from and make predictions on data by building a model from sample inputs.
Deep learning is a subset of machine learning and refers to modeling of complex relationships among layers of non-linear computational units – so-called neural networks.
Reinforcement learning solves the difficult problem of correlating immediate actions with the delayed returns they produce. The challenge here is that we know the inputs and outputs, but not quite how one led to another.
Generative Adversarial Networks are structured, probabilistic models for generating data and consist of two entities – the generator and the denominator. The denominator checks the authenticity of the data produced by the generator, whereas the generator tries to trick the denominator – it resembles trying to learn to lie without getting caught.
Transfer learning is a machine learning method where the set of learned features of a model for a specific task is reused, or repurposed, as the targeting point for a model on a second task. In practice, it’s often used for optimization.

The last two, generative adversarial networks and transfer learning bring the promise of faster progress in the field of aging. For example, algorithms can be trained in diseases with enough patients and a new understanding of biological processes could be applied to areas where data is harder to come by, such as rare diseases.

Want hear more about the future developments in Artificial Intelligence? Use the opportunity to join us for exciting talks at Data Natives Conference (November 25-26, 2019), Early Bird tickets are on sale for a limited amount of time.

What can AI solve in aging?

Long life has been praised throughout history, but life expectancy only began to increase notably in the 20th century. At the moment, the biological limit of a human body seems to be around 115 to 120 years, and scientists wonder if and how we could go even longer.

A recently published review on AI for aging explains aging as a multifactorial process far from a sudden change in isolated molecular processes or components. The human body is a complex system of connected and correlating entities – cells, tissues, organs. We are still in the early stages of discovering their causal relations and doctors still often prescribe medication on a trial and error basis.

Hence the promise of AI: it is capable of analyzing massive data sets and finding meaningful patterns, correlations, and designing models for predictions in various fields. Alex Zhavoronkov – the founder and CEO of Insilico Medicine, a Baltimore based biotech company using AI for drug development and identification of new aging biomarkers, and the co-author of the mentioned research article, says there are three significant areas how AI can be applied to aging research:

It helps construct aging clocks for age guessing.
It helps generate new chemical compounds, which are used for designing interventions to test the anti-aging related hypothesis.
It helps generate new data, by creating models in which moving one feature in time shows effects on other biological changes.

A good example illustrating the complexity of our bodies is the pipeline of novel chemistry and molecules of Insilico Medicine – with the help of AI, the company identified more than 32.000 potential targets in oncology, but only 27 are in various stages of validation. So far they found 1721 generated, profiled and scored molecules for neurodegenerative disorders, but only 9 are in various stages of validation.

As explained by Zhavoronkov, drug, molecules and biomarkers discovery is a demanding mining operation, where you fail more often than you succeed. Just think about how differently the same drug can affect two patients with the same disease. Because clinical trials are so expensive – experts need to be hired, patients enrolled, etc., – analysis supported by AI can help filter which molecules have the highest probability of success in the experimental stage. So in the future, doctors won’t have to prescribe medication on a trial and error basis again.

Aging is a complex biological process, in dire need of a better understanding. Sometimes, says Zhavoronkov, it is easier to replace the whole organ than try to identify and repair the primary causes of damage. AI promises quicker, cheaper, and more effective identification of new targets and geroprotectors (compounds that aim to affect the root cause of aging and age-related diseases).The technology also enables a different approach to clinical trials design, execution and analysis, for more efficient drug discovery.

Nobody wants to live long without vitality, good health, and potentially good looks. The focus of longevity science is not just about adding years to life but mostly about how to add life to years. Medicine is striving to offer more and more personalized treatments. While AI might not solve aging just yet, it is will be a key component in the equation.