Dementia progression subtyping and longer term disease trajectory prediction via machine learning..

Really interesting work from Faraz Faghri and co. Updated results using multi-modal data within a machine learning framework to subtype dementia and predict progression rates of cognitive decline. Accurately predicting how aggressive the disease will be over four years only using peri-diagnostic data (data collected at or around the time of diagnosis) will be extremely valuable in clinical trial recruitment going forward and also possibly rescuing failed trials. Imagine how efficient your trial could be if you specifically could target subsets of cases that reliably progressed faster or slower than others?

Research like this is important. Subtyping like this an evolution of patient stratification that made early aducanumab results from Biogen look so good. We see this as foundational to our `SPECTRUM` product we are developing that was discussed in the last blog post. Personally I’m really excited about the results and the potential growth for this research … like our friend Andy Singleton (from CARD/NIA/NIH) said “make sure you are treating the right target, in the right patients at the right time.”

Check out the paper here on biorxiv and the figure below which is a nice summary of the results and potential of this method. Now we just have to figure out what journal this goes to ???