FusionX: Automated, Standardized quantification of cell-to-cell fusion across diverse systems

Abstract

Cell-to-cell fusion, the process by which cells merge their plasma membranes to form multinucleated syncytia, is fundamental to development, physiology, and disease. Quantifying fusion in vitro typically relies on calculating the percentage of nuclei within multinucleated cells or using indirect genetic reporters. However, existing methods are laborious and error-prone, which limits standardization, accuracy, and throughput, thereby hindering meaningful mechanistic analyses. Here, we present FusionX, an AI-powered image analysis pipeline that enables automated and robust quantification of cell fusion across diverse cell types using standard membrane and nuclear dyes. FusionX integrates CellX, a fine-tuned Segment Anything Model that segments the cell boundaries of mono- and multi-nucleated cells, with Cellpose for accurate nuclear detection, enabling high-throughput and detail-rich analysis. Importantly, by extracting precise cell boundaries, FusionX provides the number of nuclei per cell together with additional single-cell parameters such as cell size and shape. Benchmarking demonstrates that FusionX delivers human-level accuracy, dramatically increases speed, and generalizes across systems, from viral fusogen-induced fusion to myogenic differentiation. By eliminating the need for specialized reporters and subjective manual quantification, FusionX paves the way for reproducible, scalable and multiparametric quantification of cell fusion in a wide range of biological contexts.