Huge thanks to our coauthors @lasgarba, Nick Fox, @FBarkhof, @jmschott, the paper reviewers, and our many wonderful colleagues working in the field!

D3PMs have provided insights across a range of conditions (Alzheimer’s and beyond) including disease biomarker timescales, quantitative support for hypothetical biomarker trajectories and disease mechanisms, disease subtype discovery, and novel strategies for clinical trials.

Pathophysiological D3PMs include network, dynamical systems and mechanistic combinations models. As with phenomenological D3PMs, each provides their own insight and utility.

Pathophysiological D3PMs model disease mechanisms that may cause pathology to ‘appear’ or ‘spread’ across the brain.

Phenomenological D3PMs include discrete, continuous, spatiotemporal, and subtyping models. Each provides their own unique insight and utility.

We provide a single framework for D3PMing that integrates phenomenological and pathophysiological models. D3PMs are directly informed by measured data, but human insight is key to the interpretability of D3PMs, avoiding the black-box nature of many AI systems.

D3PMs are an emerging set of interpretable computational tools that infer long-term disease timelines from short-term data. Our review divides D3PMs into phenomenological (higher level disease biomarker models) and pathophysiological (lower level disease mechanism models).

New paper alert! rdcu.be/dvstu Our review paper on Data-Driven Disease Progression Modelling (D3PM) was published in @NatRevNeurosci yesterday, co-led by myself, @YoungAlexL, & @fishpiechicken at @uclcs and @CmicUcl