Novel algorithms for improved detection and analysis of fluorescent signal fluctuations

Abstract

Fluorescent dyes and genetically encoded fluorescence indicators (GEFI) are common tools for visualizing concentration changes of specific ions and messenger molecules during intra-as well as intercellular communication. Using advanced imaging technologies, fluorescence indicators are a prerequisite for the analysis of physiological molecular signaling. Automated detection and avnalysis of fluorescence signals requires to overcome several challenges, including correct estimation of fluorescence fluctuations at basal concentrations of messenger molecules, detection and extraction of events themselves as well as proper segmentation of neighboring events. Moreover, event detection algorithms need to be sensitive enough to accurately capture localized and low amplitude events exhibiting a limited spatial extent. Here, we present two algorithms (PBasE and CoRoDe) for accurate baseline estimation of fluorescent detection of messenger molecules and automated detection of fluorescence fluctuations. Author summaryAnalyzing molecular signalling is crucial in understanding intra- and intercellular communication. These signals are visualized using fluorescent dyes or genetically encoded fluorescence indicators. In the brain, Ca2+ signals of glial cells are essential in deciphering complex regulatory functions in health and disease. Due to signal heterogeneity, detection and analysis are highly challenging. They can be stationary, with low amplitude and localized in cell processes, occur as prominent somatic signals or propagate as waves across cellular networks. We have developed two algorithms to analyze fluorescence transients, each tackling a specific problem. PBasE performs automatic and adaptive background correction, removing basal fluorescence fluctuations. CoRoDe automatically extracts regions of interest, explicitly including temporal information to obtain a precise segmentation, which is essential for accurate transient extraction. Combined, these algorithms are able to detect regions exhibiting low amplitude transients with small spatial extent as well as large, high amplitude signals. Extracted transients are categorized based on their peak amplitude, allowing detailed analyses by comparing changes of specific properties. In order to make these algorithms accessible, an interactive application, called Msparkles, has been designed.