Tracking without localization

We’ll be talking about two fundamental problems and their solutions.

Problem 1: The title reflects the problem that as you increase your frame rate to track increasingly fast molecules and avoid motion blur, it becomes impossible to localize molecules from the small number of photon arrivals in each frame, let alone linking localizations to form trajectories. This raises the question: how can we leverage the information contained in sparse photon arrivals in each frame to determine molecular tracks, while circumventing the localization and linking paradigm inherent to tracking? Put differently, we propose a new paradigm appropriate for molecular tracking.

Problem 2: Looking at a bright cell, with fluorescence reporting on the activity of a gene, we ask the question: what fraction of the labeled protein of interest is inherited from the mother cell versus being produced by the current cell? Answering this question immediately presents a mathematical barrier: if inherited, the amount of protein depends on the cell’s division history, turning a simple rate inference problem into a mathematically pathological one. Here, we concretely answer this question by proposing a solution through AI-assisted simulation based inference to perform inference on arbitrarily non-Markovian processes.