smBEVO: A computer vision approach to rapid baseline correction of single-molecule time series

Abstract

Single-molecule time series inform on the dynamics of molecular mechanisms that are occluded in ensemble-averaged measures. Amplitude-based methods and hidden Markov models (HMMs) frequently used for interpreting these time series require removal of low frequency drift that can be difficult to completely avoid in real world experiments. Current approaches for drift correction primarily involve either tedious manual assignment of the baseline or unsupervised frameworks such as infinite HMMs coupled with baseline nodes that are computationally expensive and unreliable. Here, we develop an image-based method for baseline correction using techniques from computer vision such as lane detection and active contours. The approach is remarkably accurate and efficient, allowing for rapid analysis of single-molecule time series contaminated with nearly any type of slow baseline drift.