NeuroSeq: A Functional Network Method for Analyzing Sets of Neural Spike Trains

A major challenge in neuroscience is understanding how dynamic ensemble-level neural activity gives rise to perception, cognition, and behavior despite the apparent irregularity of individual spike trains. Increasing evidence suggests that neural activity evolves through structured, transiently stable patterns known as metastable states, which may represent a fundamental principle of cortical computation. Here, we introduce NeuroSeq, a functional network method for analyzing sets of neural spike trains that identifies and tracks neural state transitions directly from spike-train data without assumptions about the number of states. NeuroSeq operates across three complementary scales. Applying NeuroSeq to neural recordings, we recover reproducible state transitions, recurring state families, and structured transition patterns that correspond with stimulus delivery and behavioral events. It uncovers both short-timescale state changes and long-timescale organizational structure that are difficult to detect using established approaches. These results demonstrate the utility of NeuroSeq for revealing metastable states across both short and long timescales. To facilitate broad adoption, we developed NeuroSeq GUI, a cross-platform graphical interface that provides an end-to-end pipeline from raw spike trains to interpretable neural state sequences.