Photobleaching Step Counting and Localization for Fluorescence Microscopy.

Fluorescence microscopy is a fundamental tool for studying biological systems at the single-molecule level. In particular, photobleaching step patterns can provide valuable insights into molecule interactions. Several works have proposed an automated and robust way to estimate when photobleaching steps occur. However, most methods are challenging to calibrate by non-experts in statistics or signal processing and struggle with low signal-to-noise ratios. This work introduces a supervised approach to localize photobleaching steps in fluorescent microscopy traces. Our method does not require any calibration but instead needs labels, that is, a few traces where an expert has manually provided the step positions. Our algorithm uses this information to automatically tune a change-point detection method that can reproduce the expert's annotations on new signals. We show on simulated data that our approach better copes with noise than existing ones. We then illustrate how our algorithm can be used to estimate the translation speed of a ribosome.