Synthetic Data Resource and Benchmarks for Time Cell Analysis and Detection Algorithms

Abstract

Hippocampal CA1 cells take part in reliable, time-locked activity sequences in tasks that involve an association between stimuli, in a manner that tiles the interval between the stimuli. Such cells have been termed time cells. Here we adopt a first-principles approach to comparing diverse analysis and detection algorithms for identifying time cells. We developed a resource for generating synthetic activity datasets using calcium signals recorded in vivo from mouse hippocampus using 2-photon imaging, for template response waveforms. We assigned known, ground truth values for properties of time cells in this synthetic dataset, including noise, timing imprecision, hit-trial ratio and calcium event width. These datasets were the input to a pipeline for testing multiple algorithms for time cell detection to determine the conditions for which they were best suited, and evaluate their effective operating ranges. We find that most algorithms are sensitive to noise. Only a few methods benefit from larger event widths. Reassuringly, most methods are insensitive to timing imprecision, and exhibit successful time cell detection even at low hit trial ratios. Importantly, all methods show good concordance in identifying cells as time cells.