More stuff in the paper itself! Thanks to Gabi Barbash, @aviezer_l @ramijaschek @EhudR4 @RanBalicer @LiranShlush, and everybody in the Amos Tanay lab at Weizmann. (17/17).

We are also excited to follow-up on the features our model uses to distinguish excellent longevity potential from just good potential. For example, understanding the dynamics of neutrophil/lymphocyte in the blood, or deciphering the meaning of different alk-phos levels. (16/17).

What do you want to know?.

And you can compute your own longevity score here: Our lifelong risk models, computed using only short-term history, are of great use to control for such predisposition in e.g. GWASs or predictive disease analysis. Code at (15/17).

Our models were trained on excellent data from the Clalit system, but the results are highly universal and applicable to any system (e.g., UKBB or NHANES) with minimal adjustments. You don’t need longitudinal coverage to score patients! (14/17)

BMI and glucose are markers of chronic diseases and therefore do not effectively distinguish strongly healthy individuals. (13/17)

Looking at the markers in strongly healthy patients, we observe that mildly low neutrophil counts and alkaline-phosphatase levels serve as early indicators of healthy aging. (12/17).

To validate this finding, we ran multiple tests, including comparisons with parental lifespans. (11/17)

The answer is a definitive yes! We identify patients who are healthy and devoid of any disease risk. These “strongly healthy” individuals still exhibit high variation in their longevity score. (10/17).

Now comes the fun part. We model for each individual the lifelong risk for chronic disease and their longevity potential. Naturally, the risk for disease is associated with lower longevity. Does the scoring of longevity surpass the combined sum of known disease risks? (9/17).

The distinction between high-risk and low-risk patients becomes more pronounced with increased data (i.e., more lab measurements) which becomes available during middle adulthood. (8/17)

Using our stitching approach, we can compute for each patient the lifelong probability for any disease, not limited to their longevity potential. (7/17).

The model “stitches” short-term patient histories (5 years) into a model spanning entire lifespans. When employed in the context of longevity, it distinguishes individuals with excellent and poor potential, even when their short-term survival prospects appear optimal. (6/17)

Can we observe signs of exceptional health in aging in young or middle adulthood? In particular, when considering only those patients free from known diseases? In our paper we develop a Markov model to facilitate such observations. (5/17)

So, you can be labeled healthy, and yet based on your standard lab tests, your health potential will vary significantly. Lab tests positively and negatively correlated with longevity potential include classical markers, but also some less expected indicators. (4/17)

Our exploration commences with an ML model that quantifies a remarkable spectrum of longevity potential among 80-year-olds who are in good health. (3/17)

Aging individuals are often affected by multiple diseases. And then there are the lucky ones, who enjoy robust health. But is the mere absence of known diseases sufficient for one to be classified as healthy? (2/17).

New paper from the Tanay lab on longevity is out today in @NatureAging! We employ machine learning to tackle a classical question: What makes some people live long, healthy lives, other than the fact that they are not sick with any chronic disease? (1/17).

RT @aviezer_l: So it took a lot of time – but finally our comprehensive paper on breast cancer epigenetics is out! It is a joint work of th….

And assuming we can enhance the historical normal/abnormal models for lab interpretation - perhaps this can help us to start relaxing the current dichotomy of health and disease states today in research, tomorrow in the clinic?.

In conclusion we must ask: when will physicians enjoy interpretation of “normal” lab values that takes full advantage of what modern data can offer them?.