How long can I live? This is a question that everyone thinks about but no one dares to answer, except the fortune teller on the street.
Back in 2018, Professor Steve Horvath of UCLA developed and became famous for his epigenetic-based Horvath aging clock, an “aging clock” that is important for healthcare and an aging society. importance is growing. “‘How do we age?’, we are getting close to the edge of the answer to that question.” Alex Zhavoronkov’s Leaping Aging – How Advances in Biotechnology Will Change the Global Economy ‘ wrote the book.
Previously, Shenghui reported that Deep Longevity Inc. (hereinafter referred to as “Deep Longevity”), a medical artificial intelligence company based in Hong Kong that developed an “aging clock”, received financing (for details, click to read “The advent of an AI-based aging clock, this A start-up company announced the completion of Series A financing”), the company is positioned as “a biotechnology enterprise that uses AI technology to discover biomarkers of aging”, and it has received hundreds of leading companies from international aging research such as Human Longevity Inc. million dollar financing. After the company announced the financing, there was no detailed report in the industry on the operation mechanism of the “aging clock” and the results that could be obtained. Therefore, Shenghui contacted “Deep Longevity” and conducted further discussions with it.
“The emergence of the aging clock is a fusion of technology and trends”
“Deep Longevity” was hatched from Insilico Medicine, the leading artificial intelligence drug research and development company in China. At present, the company has 9 people. The chief scientist is Polina Mamoshina, a doctor of computer science from Oxford University. The founder, Alex Zhavoronkov, is also the founder and CEO of Intech Intelligence. Alex told Shenghui that the reason why “Deep Longevity” was spun off from Inke Intelligence is because “Inke Intelligence focuses on preclinical drug discovery and artificial intelligence innovative drugs, while ‘Deep Longevity’ has a completely different target market. and expertise, there is little overlap, so we decided to divest it to unlock its value quickly.”
The company has been established for less than three years. According to Alex, the company received its first offer not long ago, and some listed companies have begun to seek the acquisition of “Deep Longevity”.
Alex graduated from Johns Hopkins University, majoring in both AI and medicine. Since 2011, he has published more than 120 cutting-edge scientific research results and 2 scientific research books. Serial entrepreneur, worked as an engineer at Canadian ATI Technology, which was acquired by AMD for $9 billion.
Talking about the source of inspiration for the “aging clock”, Alex said that he tried to use data from patients with cancer, diabetes, Alzheimer’s, etc. Likewise, there is hardly a single unifying variable to do the basic quality contral of deep neural networks, and he quickly realized that the only universal unifying feature that is always present in everyone is age.
Most people’s perception of age is that age is the number of birthdays, but scientists believe that age does not predict how long we will live, and the gap between the popular age and the actual biological age may be more than 30 years.
How to know your biological age and predict how long you can live? “Deep Longevity” wants to predict your lifespan or even reverse your age through AI. The company trained a set of deep neural networks based on blood test data, microbial data and other data. When there is a large number of sample data, the deep neural network learns human biology and is interpretable, so that it can find new Aging biomarkers and establishing causal relationships. “On the one hand, we use AI to find new ways to analyze longitudinal data, and on the other hand, we use longitudinal data to find new ways to build better AI,” Alex said.
Figure | A related recent paper by Alex and Sinovation and their AI Institute describes the rationale
In addition to the above-mentioned integration of AI technology represented by deep neural networks and the rapid development of blood testing and other technologies in recent years, Alex believes that the emergence of “aging clock” is also a fusion of trends.
The advent of the Horvath aging clock brought renewed focus to aging research, although it was not very popular among Big Pharma at the time. But in the past two years, longevity biotechnology has gradually evolved into an industry, and some “longevity drugs” such as senolytics, NAD+ boosters, metformin, rapalogs, etc. have appeared, and even venture funds and investors have begun to pay attention to and invest in this industry. Companies in the field, Google, Apple, Facebook, Amazon and other giants, needless to say, have also invested a lot of money and energy in trying to discover technology that prolongs life.
Alex believes that the main competitiveness of “Deep Longevity” comes from its first-mover advantage, because there are not many players in this field at present. The company has immediate access to key researchers and scholars in the field of aging research, which can help them continue to develop new biomarkers and interventions.
Talking about the company’s vision, Alex said: “Our ultimate vision is to create an ecosystem of our aging clock, clinical centers and insurance companies working together to make humans live longer and keep the best possible life. Optimum performance. Instead of limiting human age to a reference range, we try to keep people in the best physical condition possible – 20 to 40 years old.
Looking for a Chinese partner
How does the “aging clock” of “deep longevity” work? How accurate is it?
Alex said their search for biomarkers of aging used millions of samples from large population databases, such as a study called NHANES in the United States, the UK Biobank in the United Kingdom, and many others. Longitudinal data. “Deep Longevity” inputs a large amount of characteristic data to predict age or health status, and then uses a new, independent database to test and tune the deep neural network and check its accuracy.
Often, different data types and different data qualities produce different error rates, Alex said. For example, blood-based aging clocks have a 5-year mean absolute error, which means that if you are healthy, the predicted age will be five years older or younger than your true age. If the patient is healthy, the data error rate for methylation and transcription data is lower. They also developed an algorithm to explain existing results, such as what makes you younger or older.
Figure | What are the main data types used to predict age? (Source: Shenghui)
According to Alex, the prediction process will start with simple starting points, such as photos, activities, heart rate and questionnaires, and then can enter historical blood test data, such as cholesterol, inflammatory markers, hemoglobin concentration, etc., if there is a previous If the data is retained, there is no need to go to a specific laboratory for additional testing until this step. The “aging clock” will then make predictions based on gene expression, protein expression, microbes, and more recently they are developing an aging clock for mental age.
For end users, they report seeing where in their body they are aging faster and adjust their behavior or intervene appropriately to reduce aging. “The picture is more intuitive to explain,” Alex said, and the following is a screenshot of a report that users got:
Figure | The user only filled in a part of the information, so the data is not of much value, and it can only be concluded that the user needs more sleep to improve blood aging. Because the “blood age” is beyond the age shown in the photo. The system gives a weighted average of 41 and also points for improvement.
The company is primarily working with medical clinics in the US and UK, and has also launched an app and website-based biomarker tool for aging in the Western Hemisphere. As for the Chinese market, because of data rights issues, “Deep Longevity” needs to license its technology and software to a mainland partner in order to have the right to use data. Currently they are looking for a partner.
In terms of approval and regulation, Alex responded that these aging clocks are not currently constrained because aging is currently not considered a disease. Regarding the data security issues of public concern, he replied that he would not collect sensitive private information of users, nor would he be able to identify a person based on the information entered by users. He suggested that users use nicknames instead of real names when uploading pictures and information.
“In China, aging research and longevity biotechnology are still in their infancy, and we are organizing the world’s largest online conference on aging research and drug discovery. Scholars and experts in the field will present their research, but very few of them come from Speakers and representatives from mainland China,” Alex said.