Abstract
When we get sick, we want to be resilient and recover our original health. To measure resilience, we need to quantify a host's position along its disease trajectory. Here we present Looper, a computational method to analyze longitudinally gathered datasets and identify gene pairs that form looping trajectories when plotted in the space described by these phases. These loops enable us to track where patients lie on a typical trajectory back to health. We analyzed two publicly available, longitudinal human microarray datasets that describe self-resolving immune responses. Looper identified looping gene pairs expressed by human donor monocytes stimulated by immune elicitors, and in YF17D-vaccinated individuals. Using loops derived from training data, we found that we could predict the time of perturbation in withheld test samples with accuracies of 94% in the human monocyte data, and 65-83% within the same cohort and in two independent cohorts of YF17D vaccinated individuals. We suggest that Looper will be useful in building maps of resilient immune processes across organisms.