From Biological Aging Clocks to Longevity Medicine: An Artificial Intelligence Odyssey
Ramkumar Hariharan
Data Science Engineering Faculty,
Program Director and Senior Scientist
Source: myWMAnews
Isn’t aging the ultimate “Emperor of All Maladies ?
Immune Function
Liver Function
Kidney Function
Cognition
Cancer
AD
CVD
Infections
Source : Seals DR & Melov S, 2014, Aging
Taeuber paradox, Estimates by Matt Kaeberlein, UW
Longevity – observations
Jeanne Calmet, lived to 122
Jonathan, 189
Albert Einstein, lived to 76
Clams, 400+
What will it take to live longer (No pun intended) ?
source alex Zharanov, Nature Aging, 2021
Of biomarkers and disease…
Fever
Source: Harvard health, unsplash,Yale medicine,
Cardiovascular health
Lung function, Cardio health
Diabetes
Aging
When Serendipity Met Scientific Brilliance
Epigenome sits on the genome & Directs how information in decoded
Source: cuinnedits.com, frontiersin.org
Horvath Clocks, DNAm Age & ML
353 CpG (features)
Elasticnet Regression (Lasso + Ridge) -> ~ 21,000 CpG (features) ->
Error 3.6 yrs
DNAm age Vs. Chrono age corr = 0.97
Total N ~ 8000 from 82
”normal” datasets
• Train = 40, Valid = 30
• 51 cell types & tissues
• Cancers DNAm +
• iPSCs DNAm = 0 • Breast Tissue DNAm +
•
We
Rooted in biology !!!
need multi-omic, causal clocks
Hallmarks of Aging
• Carlos López-Otín et al, 2013
Hallmarks of Aging
• Carlos López-Otín et al, 2023
Hallmarks of Aging: An expanding Universe
Source: Carlos López-Otín et al, 2023
AI for drug discovery and personalization
Which one, when, how much and how long
amazon.com, agelessforever.net,
healthsite.com
Source:
Lifeextension.com,
Clinical trials for first generation longevity medicine
• Pick one or more gold standard drugs
• Validate them on aging clock
• Build causal aging clocks
•
2023 and beyond
Personal Research Goals & Collab Request
Multi-omic
in cancer survivors & Down Syndrome patients
clocks
Building multi–omic ML powered tools, Biology–derived feature reduction * Rxn 1 Rxn 3 Rxn 2 Rxn 4 Rxn 5 Rxn 6 Rxn 7 Striatum Cortex Liver Adipose brown Adipose white gonadal Brainstem Cerebellum Corpus Callosum Cortex Gastrocnemius Heart Hippocampus Hypothal/Thalamus Liver Skin Striatum Q175 Q20 Q80 Q92 Q111 Q140 2 mo 6 mo 10 mo Q175 Q7 Differential expression Tissue specificity Identify differentially expressed pathways across age and Q-length Map metabolic pathway topology; populate with expression data PathWave metabolic network analysis oPOSSUM TF enrichment in metabolic genes Metabolic gene Rxn 1 Rxn 3 Rxn 2 Rxn 4 Rxn 5 Rxn 6 Rxn 7 Wildtype Mutant CAG length dependent TFs 6 mo AILoT Artificial Intelligence for Longevity Toolbox
Immunotherapy Effectiveness in Cancer in relation to biological age
Acknowledgements
Ricardo–Baeza Yates, Professor of the Practice
Usama Fayyad, Executive. Director
Pramod Nagare, Senior Data Engineer
Many others….
Jaren Walker, Director of Partnerships, Seattle
Dave Thurman, Dean and CEO, Seattle
Senior Leadership team @ Boston campus
João Pedro de Magalhães
