
Humanity Co-founder Michael Geer on using machine learning to bring better health outcomes to everybody – and why building models in the open speeds adoption.
Yesterday, we discussed research that has demonstrated that an estimation of biological age can be made from any blood test panel, meaning that an accurate, scalable, affordable test of biological age will be available to pretty much anyone who wants one, something that could have enormous impact for preventative healthcare and people taking charge of their own longevity trajectory.
Longevity.Technology: While the publication has not yet been peer-reviewed and is still in preprint, we were keen to find out more straightaway, so we reached out to Michael Geer, Co-founder of Humanity, an app that monitors your rate of aging, and one of the paper’s authors and sat down with him to find out more about platform and what it could mean for longevity.
Michael Geer on…
Inspired by blood
The UK Biobank has about 300,000 participants’ records of blood biomarkers,and we asked ourselves: “Can we make the power of this much more accessible to the mainstream?” This was something Humanity had already done with digital biomarkers, so we started on applying AI machine learning to get to something much more robust than has existed in the past.
We were able to improve on past blood models, getting a 9% improvement on biological age prediction – that in itself would be enough for a big announcement, but the second breakthrough is, we feel, a lot more important. Any blood biomarkers that a person has had tested, for whatever reason, they will be able to upload that data and if that data contains only 5 or 6 of the 60 UK biobank-held markers that we analyzed, our model will be able to give them a high accuracy measurement of biological age. We’re bringing biological age tracking to the masses.
Sharing the knowledge
We want to be very open with these models; we’re not slapping “proprietary” on top of them because it’s important that people know how the models were constructed, agree with them (although there’s never complete consensus in science!) and they are done on an open dataset – the UK Biobank. These things are all very important to us, and we would love comments and to talk to people who would like to review the paper.
The unique position that we thought we could bring to the space was to work with everybody to get more stuff out to more people. This technique – and we would love to see people add to it – it’s about having great science and bringing it to the masses. We’d love to see it reverberate in the sector.
The significance of biological age as a key biomarker leads to the importance of everything being open; there needs to be open data sets people can access and the methodology needs to be quite open, as written up in this paper. I think that brings the belief – there are conversations being had about which biomarkers are good or bad, but it’s hard to argue about data. We expect to have discussions around the data and the methodology, but as it’s all open, we do believe this can be accepted by medical practitioners.