Skip to main content
. 2024 Jan 25;19(2):285–298. doi: 10.1016/j.stemcr.2023.12.008

Table 1.

Feature names and descriptions

Feature Feature class Description
MIB Burst Mean interburst interval
MFT Burst Mean time from the peak to the end of a burst
GEC Graph Global efficiency calculated on cross-correlogram (CCG)-inferred connectivity graph
DEC Graph Network density based on CCG graph
CVI Spike time Coefficient of variation of the interspike interval
MIS Spike time Mean interspike interval
RF Time series Peak frequency of Fourier power spectrum
RM Time series Magnitude of the corresponding RF feature
EAF Time series First 1/e crossing of autocorrelation function
AMI Time series Automutual information
SFR Time series Proportion of slower timescale fluctuations that scale with linearly rescaled range fits
LPF Time series Total power in the lowest fifth of frequencies in the Fourier power spectrum
MEF Time series Mean error from a rolling 3-sample mean forecasting
PAM Time series Longest period of consecutive values above the mean
EFD Time series Exponential fit to successive distances in two-dimensional embedding space
CCD Time series Change in correlation length after iterative differencing
TCT Time series Trace of covariance of transition matrix
SES Time series Shannon entropy
CFS Time series Centroid of the Fourier power spectrum
SFD Time series Proportion of slower timescale fluctuations that scale with detrended fluctuation analysis
MD5 Time series Mode of Z scored distribution (5-bin histogram)
MD10 Time series Mode of Z scored distribution (10-bin histogram)
FMA Time series First minimum of autocorrelation function
FMI Time series First minimum of the automutual information function
TRS Time series Time-reversibility statistic
TEA Time series Time intervals between successive extreme events above the mean
RFT Time series Exponential fit of successive peak frequencies in the Fourier power spectrum
PDE Time series Proportion of successive differences exceeding 0.04 SD

Subset of DeePhys features and their respective abbreviations as used throughout the paper. Time series features that were inferred at the single-cell or the network level are denoted by a leading s or n, respectively (e.g., sRM refers to the single-cell regularity magnitude). See Table S1 for a complete list of features.