Facebook PixelIn what ways can AI be used to improve longevity research and/ or medicine?
Create newCreate new
Sessions onlySessions only
Ideas onlyIdeas only
Brainstorming session

In what ways can AI be used to improve longevity research and/ or medicine?

Image credit: Photos by Markus Spiske and Pixabay from Pexels

Shubhankar Kulkarni
Shubhankar Kulkarni Jan 22, 2021
Recently, advances in deep learning have enabled the development of AI systems that not only compete but outperform humans in many tasks. AI has also empowered researchers and physicians with new tools.

What are the different ways in which AI can be used in longevity research and longevity medicine?
Creative contributions

Deep Aging Clocks that estimate an individual's biological age

Shubhankar Kulkarni
Shubhankar Kulkarni Jan 22, 2021
There is little consensus on the method that can be used to define human "biological age" that predicts mortality, diseases, or frailty and it can be largely different from the human chronological age. Biological age is variable and may change in response to anti-aging or longevity-enhancing interventions.

There are different types of Deep Aging Clocks (DACs):
  1. Deep Blood Biochemistry and Cell-Count Aging Clocks: The first deep aging clock published by Zhavoronkov in 2016 utilized deep neural networks (DNNs). Using blood tests, the team generated a dataset of over 60,000 healthy subjects. DNNs outperformed the machine learning models in every test. In another study, population differences in the aging clocks were identified by three trained DNNs. As an example, "testing Korean and Eastern European data with a DNN trained on Canadian data revealed that Koreans on average appeared younger than their chronological age".
  2. Deep Imaging Aging Clocks: A research team at Haut.AI exploited photoimaging. The deep photographic aging clock uses the images of the corners of the eye and can predict the age of an individual within an accuracy of 1.9 years mean absolute error.
  3. Deep Transcriptomic Aging Clocks: In 2018, the first transcriptomic aging clock developed using deep and other machine learning techniques based on gene expression data from muscle tissue was published. It developed was in which genes could be prioritized as possible targets for pharmaceutical intervention in sarcopenia and other muscle-wasting diseases.
  4. Other Data Types: Wearable devices and mobile phones provide huge amounts of biologically relevant data. In 2018, age-associated changes in physical activity were studied to predict age using neural networks. A deep learning-based model trained on activity-monitor data predicted age with high accuracy.

Using AI-powered DACs, physicians may be able to more precisely assess and monitor an individual's health and associated risks and provide appropriate treatment. Some researchers think that DACs should become an essential part of the physician diagnostic kit.

[1]Zhavoronkov, A., Bischof, E. & Lee, KF. Artificial intelligence in longevity medicine. Nat Aging 1, 5–7 (2021). https://doi.org/10.1038/s43587-020-00020-4

[2]Zhavoronkov A, Mamoshina P. Deep Aging Clocks: The Emergence of AI-Based Biomarkers of Aging and Longevity. Trends Pharmacol Sci. 2019 Aug;40(8):546-549. doi: 10.1016/j.tips.2019.05.004. Epub 2019 Jul 3. PMID: 31279569.

Deep algorithms for early detection of diseases

Shubhankar Kulkarni
Shubhankar Kulkarni Jan 28, 2021
Recently, a group performed a 3-year precision medicine study on a cohort of 1190 adult participants. They sequenced the whole genome of the participants and supplemented it using metabolomics, advanced imaging, clinical tests, and family and medical history. The utilization of these supplementary data is known as deep phenotyping. Only using the genome sequence revealed that 206 (17.3%) of the 1190 participants had at least one pathogenic or likely pathogenic genetic variant suggesting a genetic risk for disease. Genotype and phenotype associations revealed that only 137 (11.5%) participants had a risk (of diseases like dyslipidemia, cardiovascular diseases, and endocrine diseases). When deep learning algorithms were used to assess the associations between genomics and metabolomics, it identified 61 (5.1%) heterozygotes with phenotypic manifestations of diseases precursors that affected their serum metabolite levels of parameters involved in amino acid, lipid, and vitamin pathways. Deep learning algorithms could correctly identify and more precisely estimate the risk of diseases in the participants.

[1]Ying-Chen Claire Hou, Hung-Chun Yu, Rick Martin, Elizabeth T. Cirulli, Natalie M. Schenker-Ahmed, Michael Hicks, Isaac V. Cohen, Thomas J. Jönsson, Robyn Heister, Lori Napier, Christine Leon Swisher, Saints Dominguez, Haibao Tang, Weizhong Li, Bradley A. Perkins, Jaime Barea, Christina Rybak, Emily Smith, Keegan Duchicela, Michael Doney, Pamila Brar, Nathaniel Hernandez, Ewen F. Kirkness, Andrew M. Kahn, J. Craig Venter, David S. Karow, C. Thomas Caskey. Precision medicine integrating whole-genome sequencing, comprehensive metabolomics, and advanced imaging. Proceedings of the National Academy of Sciences Feb 2020, 117 (6) 3053-3062; DOI: 10.1073/pnas.1909378117

Machine learning-powered molecular docking tools to predict the side-effects of novel anti-aging therapies

Shubhankar Kulkarni
Shubhankar Kulkarni Jan 29, 2021
Most drugs have side-effects. Even then the search for newer drugs is on the rise. Drug trials are a time-consuming procedure, and it takes years for a drug to become approved for use. Most current side-effect prediction methods focus on the drug's chemical and biological properties. Therefore, machine learning modeling methodologies are a good alternative.

Recently, "a method that uses the drug-drug interactions from DrugBank, drug-drug interactions from the network, single nucleotide polymorphisms, and side effect anatomical hierarchy as well as chemical structures, indications, and targets" was used to predict the side-effects of a drug. The method used these known associations between drugs and their side-effects and drug-drug interactions and molecular similarities between drugs to associate the new drug with its potential side-effects. The performance of this new method was found to be 3.5% better than that obtained by only the target, chemical, and indication features.

Other machine learning-powered methods to predict drug side-effects are:
  1. A Novel Triple Matrix Factorization Method for Detecting Drug-Side Effect Association Based on Kernel Target Alignment
  2. Prediction of Drug Side Effects with a Refined Negative Sample Selection Strategy
Thank you @Nitish for the idea.

[1]Sukyung Seo, Taekeon Lee, Mi-hyun Kim, Youngmi Yoon, "Prediction of Side Effects Using Comprehensive Similarity Measures", BioMed Research International, vol. 2020, Article ID 1357630, 10 pages, 2020. https://doi.org/10.1155/2020/1357630

[2]Xiaoyi Guo, Wei Zhou, Yan Yu, Yijie Ding, Jijun Tang, Fei Guo, "A Novel Triple Matrix Factorization Method for Detecting Drug-Side Effect Association Based on Kernel Target Alignment", BioMed Research International, vol. 2020, Article ID 4675395, 11 pages, 2020. https://doi.org/10.1155/2020/4675395

[3]Haiyan Liang, Lei Chen, Xian Zhao, Xiaolin Zhang, "Prediction of Drug Side Effects with a Refined Negative Sample Selection Strategy", Computational and Mathematical Methods in Medicine, vol. 2020, Article ID 1573543, 16 pages, 2020. https://doi.org/10.1155/2020/1573543

An integrative multifactorial AI-powered aging metric

Juran Jan 31, 2021
Thank you for starting this cool topic! I enjoyed reading your contributions because it's something I didn't know existed in a physical form. I thought it was still just an idea.

By searching the web, I found that someone does what you explained in a Deep Aging Clocks contribution, but in a very sophisticated integrative form.

You mentioned few aging clocks based on different sources of data (Deep Blood Biochemistry, Cell-Count Aging Clocks, Deep Imaging Aging Clocks, and Deep Transcriptomics Aging Clocks), but what the
Deep Longevity company does is the integration of various multifactorial aging clocks. From their
Biological Age Report example, you can see that they calculate various aging clocks using the AI-tools (hematological, gut microbiome, heart-rate, methylation, photo (face+neck), blood-transcriptome, behavior (lifestyle, habits), all enriched with personal medical history) and combine them to get a universal "AgeMetric" .

The company also developed an app that gives you an overview of your multifactorial longevity clock results, gives you easy-to-follow tips on how to prolong your life by changing habits, compares the old reports with the new ones, and help you keep track of your progress. It also enables a simple data input by taking a photo of a blood test or a selfie.

By integration of all these AI-powered metrics, I believe we are getting closer to a personalized track-and-react longevity tool that will help us live longer.

[1]Mamoshina, P., Kochetov, K., Putin, E., Cortese, F., Aliper, A., Lee, W.-S., Ahn, S.-M., Uhn, L., Skjodt, N., Kovalchuk, O., et al. (2018). Population Specific Biomarkers of Human Aging: A Big Data Study Using South Korean, Canadian, and Eastern European Patient Populations. Journals Gerontol. Ser. A 73, 1482–1490.

Shubhankar Kulkarni
Shubhankar Kulkarni3 months ago
That is a great find Juran K. ! Thank you for sharing.

Add your creative contribution

0 / 200

Added via the text editor

Sign up or


Guest sign up

* Indicates a required field

By using this platform you agree to our terms of service and privacy policy.

General comments