Finally, we use FinRegistry (8,000+ features) + deep-learning to predict 1-year mortality for the entire Finnish population (AUC=0.95)! Derive a powerful digital marker of aging And show predictions are subject to geographical and socio-economic unfairness www.nature.com/articles/s43...

Deep learning-based prediction of one-year mortality in Finland is an accurate but unfair aging marker - Nature Aging

Here the authors show that deep learning accurately predicts one-year mortality using nationwide Finnish data. Despite robust performance and potential as an aging marker, fairness analyses reveal pre...

Second, we show that the variability in polygenic score performance is larger between biobanks than between the methods used to build the scores t.co/ESmOPB8vr9

Evaluation of polygenic scoring methods in five biobanks shows larger variation between biobanks than methods and finds benefits of ensemble learning

Methods of estimating polygenic scores (PGSs) from genome-wide association studies are increasingly utilized. However, independent method evaluation i…

Then a couple of papers on Polygenic Scores. First, we provide country-specific cumulative incidence estimates for 18 diseases by PGS levels. rdcu.be/dKAK4

A unified framework for estimating country-specific cumulative incidence for 18 diseases stratified by polygenic risk

Nature Communications - Here the authors present a framework for estimating disease risk using PGS accounting for country, age and sex. They find that PGSs have a significant sex-specific effect on...

very nice to see such a mature use of nation-wide Dutch health data!

Thanks Michel!!

Congrats Michel!! let’s catch up in the next months

so no matter which genetic risk factor you look at, probably very little predictive of our definition of progression

we didn’t explore that, but it is a good suggestion. basically you want to look how the genetics of risk factors of susceptibility associate with survival. one think to consider is that disease-specific survival is very little heritable in our study.

There are some examples for other diseases as well, but generally very few (compared to susceptibility) and not always replicable.

Under the current definition, genetic signals for disease progression will be hard to detect and existing disease PGS are not very relevant for progression. One can come up with better definitions of disease progression, but would they be easily scaled across biobanks?