TPE: finding the turning point in disease progression

A disease does not progress in a straight line. It drifts, then turns, and the turn is the part that matters.

A biomarker can changes gradually until it passes a critical point and then accelerates. The trajectory bends, and from one patient to the next that bend lands in a different place, pushed around by age, genetics, and everything else that makes a person who they are. Average the curves together and the turn smears into nothing.

I built the transition point estimator (TPE) to find that turn directly. It uses machine learning to model the nonlinear link between a biomarker and an outcome, keeps the confounders in check, and reads off the point where each marker’s behavior shifts. That point becomes its cutpoint.

A cutpoint set this way is important in how the disease actually moves. Find the turn, and you find the moment worth acting on.