Partaker: deep-learning-based single-cell-resolution analysis of multi-dimensional, long-term, microfluidic-based time-lapse microscopy data

Abstract

Abstract Motivation Deep-learning segmentation models for microbial time-lapse fluorescence microscopy already exist, but they are often difficult to use consistently across experiments and are not packaged with unified workflows for combining models, quantifying multi-channel fluorescence, and scaling analyses to large microfluidic imaging datasets. Results To address these challenges, we developed Partaker, an easy-to-use Python-based graphical tool for deep-learning segmentation, multi-channel fluorescence quantification, and morphological analysis of microbial cells over time. Partaker supports extensible model integration, efficient handling of large imaging datasets, and interactive visualization, enabling time-resolved analysis of population fluorescence distributions from per-cell measurements. We demonstrate performance using a two-strain validation experiment (housekeeping and inducible) and show robust recovery of population-level fluorescence dynamics with single-cell resolution. Availability and implementation Partaker is implemented in Python and is available on GitHub (https://github.com/SamOliveiraLab/partaker). A versioned release of the software is archived on Zenodo (https://doi.org/10.5281/zenodo.18844425). Documentation, example datasets, and installation instructions are provided in the repository.