Collective flow of circadian clock information in honeybee colonies (Data)

This repository contains the data used in the paper "Collective flow of circadian clock information in honeybee colonies".

Paper: Collective flow of circadian clock information in honeybee colonies

Code and more details are provided in the README file of speedtransfer repository.

Description of included files

All data sets exist for the period 01.08.-25.08.2016 and 20.08-14.09.2019.

mean_velocity_2016.csv and mean_velocity_2019.csv

The mean velocity for each bee and age is averaged over 10-minute time windows.

Keys:

• velocity: Mean euclidean distance of two consecutive points of a bee's hive position.

• datetime: Date in year-month-day hour:minute:seconds+ms:ns format.

• age: Age in days. Can be NaN if the bee has no associated death_date.

velocity_2088_2019.csv and velocity_5101_2019.csv

The movement speed [mm/s] of two individual bees with the bee id 2088 and the bee id 5101 for the period 2019.

Keys:

• velocity: Euclidean distance of two consecutive points of a bee's hive position.

• time_passed: Time [s] in between the datetime of that current and the last previous detection.

• datetime: Date in year-month-day hour:minute:seconds+ms:ns format.

cosinor_2016.csv and cosinor_2019.csv

A cosinor fit of the velocity per bee for a time window of 3 consecutive days according to the method proposed by Cornelissen.

Keys:

• mesor: Rhythm-adjusted mean of a cosine with a period of one day fitted to the velocity data of the individual over three days. Can be NaN if the fit did not converge due to a lack of data points.

• amplitude: Amplitude of a cosine with a period of one day fitted to the velocity data of the individual over three days. Can be NaN if the fit did not converge due to a lack of data points.

• phase: Acrophase of a cosine with a period of one day fitted to the velocity data of the individual over three days. Can be NaN if the fit did not converge due to a lack of data points.

• p_value: P-value of an F-test for overall significance of a cosinor fit with a period of one day fitted to the velocity data of the individual over three days. Can be NaN if the fit did not converge due to a lack of data points.

• p_mesor: P-value of the mesor coefficient of a cosinor fit with a period of one day fitted to the velocity data of the individual over three days. Can be NaN if the fit did not converge due to a lack of data points.

• p_amplitude: P-value of the amplitude coefficient of a cosinor fit with a period of one day fitted to the velocity data of the individual over three days. Can be NaN if the fit did not converge due to a lack of data points.

• p_acrophase: P-value of the phase coefficient of a cosinor fit with a period of one day fitted to the velocity data of the individual over three days. Can be NaN if the fit did not converge due to a lack of data points.

• p_reject: P-value of a F-test for model validity.

• r_squared: R² value of a cosinor fit with a period of one day fitted to the velocity data of the individual over three days. Can be NaN if the fit did not converge due to a lack of data points.

• r_squared_adj: Adjusted R² value of a cosinor fit with a period of one day fitted to the velocity data of the individual over three days. Can be NaN if the fit did not converge due to a lack of data points.

• p_ks: P-value of a Kalgomorov-Smirnoff test of residual normality.

• p_hom: P-Value of F-test of variance homogeneity of cosinor fit.

• dw: Durbin-Watson statistic for the independence of the residuals.

• p_runs: P-value of runs test of independence of residuals.

• RSS: Residual sum of squared - the sum of squared differences between the data and the estimated values from the fitted model

• SSPE: Pure error sum of squares of cosinor fit.

• bee_id: Numeric unique identifier per individual bee.

• age: Age in days. Can be NaN if the bee has no associated death_date.

• date: Date and time in year-month-day hour:minute:seconds+ms:ns format. The hour is always 12.

• n_data_points: Number of data points per cosinor fit.

• data_point_dist_max: Maximum temporal distance between two consecutive timepoints of velocity data.

• data_point_dist_min: Minimum temporal distance between two consecutive timepoints of velocity data.

• data_point_dist_mean: Mean temporal distance between two consecutive timepoints of velocity data.

• data_point_dist_median: Median temporal distance between two consecutive timepoints of velocity data.

• day_mean: Mean velocity during the daytime defined as the time between 9 and 18 o'clock.

• day_std: Standard deviation of velocity during the daytime defined as the time between 9 and 18 o'clock.

• night_mean: Mean velocity during the nighttime defined as the time between 21 and 6 o'clock.

• night_std: Standard deviation of velocity during the nighttime defined as time between 21 and 6 o'clock.

• ad_fuller: P-value of augmented Dickey-Fuller test for testing whether the velocity data is stationary.

• fit_type: Median time bin in seconds used for fit, e.g. 3600 means that the median over a time window of 3600s is used for the fit.

• ci_acrophase_lower: Confidence interval lower bound for the acrophase fit value.

• ci_acrophase_upper: Confidence interval upper bound for the acrophase fit value.

• ci_mesor_lower: Confidence interval lower bound for the mesor fit value.

• ci_mesor_upper: Confidence interval upper bound for the mesor fit value.

• ci_amplitude_lower: Confidence interval lower bound for the amplitude fit value.

• ci_amplitude_upper: Confidence interval upper bound for the amplitude fit value.

interactions_side0_2016.csv, interactions_side0_2019.csv and interactions_side1_2016.csv, interactions_side1_2019.csv

The bee interactions and their post-interaction velocity change. An interaction between two bees (bee0 and bee1) is defined when two bees are detected simultaneously in the hive with a confidence threshold of 0.25, the distance between the markings on their thorax bodies is no more than 14 mm. These interactions are combined into one interaction if the same detections occur within a time interval of 1 second or less between them. The resulting interaction data frames per bee are concatenated with the estimates of the cosinor dataframe.

Keys:

• bee_id0: Numeric unique identifier per individual bee.

• bee_id1: Numeric unique identifier per individual bee.

• interaction_start: Timestamp indicating interaction start time point.

• interaction_end: Timestamp indicating interaction end time point.

• x_pos_start_bee0: Numeric x-position of bee relative to hive at interaction start.

• y_pos_start_bee0: Numeric y-position of bee relative to hive at interaction start.

• theta_start_bee0: Numeric angle of bee relative to hive at interaction start.

• x_pos_start_bee1: Numeric x-position of bee relative to hive at interaction start.

• y_pos_start_bee1: Numeric y-position of bee relative to hive at interaction start.

• theta_start_bee1: Numeric angle of bee relative to hive at interaction start.

• x_pos_end_bee0: Numeric x-position of bee relative to hive at interaction end.

• y_pos_end_bee0: Numeric y-position of bee relative to hive at interaction end.

• theta_end_bee0: Numeric angle of bee relative to hive at interaction end.

• x_pos_end_bee1: Numeric x-position of bee relative to hive at interaction end.

• y_pos_end_bee1: Numeric y-position of bee relative to hive at interaction end.

• theta_end_bee1: Numeric angle of bee relative to hive at interaction end.

• vel_change_bee0: Numeric post-interaction absolute change of velocity: abs = vafter-vbefore with vbefore and vafter are calculated as the mean velocity 30s before and after the interaction.

• rel_change_bee0: Numeric post-interaction relative change of velocity: rel = (vafter-vbefore)/vbefore with vbefore and vafter are calculated as the mean velocity 30s before and after the interaction.

• vel_change_bee1: Numeric post-interaction absolute change of velocity: abs = vafter-vbefore with vbefore and vafter are calculated as the mean velocity 30s before and after the interaction.

• rel_change_bee1:  Numeric post-interaction relative change of velocity:rel = (vafter-vbefore)/vbefore with vbefore and vafter are calculated as the mean velocity 30s before and after the interaction.

Modeling bees as rectangular mask to determine if bees body overlap when interacting - see more details in speedtransfer repository:

• x_trans_focal_bee0: Translated and rotated x-position relative to hive.

• y_trans_focal_bee0: Translated and rotated y-position relative to hive.

• theta_trans_focal_bee0: Translated and rotated theta relative to hive.

• x_trans_focal_bee1: Translated and rotated x-position relative to hive.

• y_trans_focal_bee1: Translated and rotated y-position relative to hive.

• theta_trans_focal_bee1: Translated and rotated theta relative to hive.

• overlapping: Bool indicating whether rectangular masks modeling the body of bees overlap.

Cosinor fit parameters - see more detailed in Cosinor data frame:

• amplitude_bee0: Amplitude of the cosinor fit per bee of the day of the interaction.

• phase_bee0: Phase of the cosinor fit per bee of the day of the interaction.

• p_value_bee0: P-value of the cosinor fit per bee of the day of the interaction.

• r_squared_bee0: R² value of the cosinor fit per bee of the day of the interaction.

• amplitude_bee1: Amplitude of the cosinor fit per bee of the day of the interaction.

• phase_bee1: Phase of the cosinor fit per bee of the day of the interaction.

• p_value_bee1: P-value of the cosinor fit per bee of the day of the interaction.

• r_squared_bee1: R² of the cosinor fit per bee of the day of the interaction.

interactions_side0_null_model_2016.csv and interactions_side0_null_model_2019.csv

A null model for bee interactions and their post-interaction velocity change. The interaction null model is created by taking the distribution of the start and end times of a given interaction dataframe and selecting two random bees at those times that the bees were detected in the hive at that time. These pairs of bees are considered as "interacting" and their post-interaction speed change is calculated. The resulting interaction data frames per bee are concatenated with the estimates of the cosinor dataframe. The position data is relative to pixels not to the hive coordinates.

As this is a null model the following keys are the same as described in the Interaction data frame.

Keys:

• bee_id0

• bee_id1

• interaction_start

• interaction_end

• x_pos_start_bee0

• y_pos_start_bee0

• theta_start_bee0

• x_pos_start_bee1

• y_pos_start_bee1

• theta_start_bee1

• x_pos_end_bee0

• y_pos_end_bee0

• theta_end_bee0

• x_pos_end_bee1

• y_pos_end_bee1

• theta_end_bee1

• vel_change_bee0

• rel_change_bee0

• vel_change_bee1

• rel_change_bee1

• age_bee0

• phase_bee0

• amplitude_bee0

• r_squared_bee0

• p_value_bee0

• age_bee1

• amplitude_bee1

• r_squared_bee1

• p_value_bee1

• phase_bee1

interaction_tree_paths_2016.csv and interactions_tree_paths_2019.csv

Graph-theoretic interaction tree paths. By tracing back interactions that positively influenced the speed of a focal bee, we constructed a graph-theoretic tree structure starting from a young rhythmic bee and recursively adding activating (velocity change parent > 0) individuals. We examined the impact of sequential interactions among bees occurring between 10 am and 3 pm, focusing on a subgroup of n = 1000 bees that are significantly rhythmic, younger than 5 days old, and peak in activity after 12 pm. We limited the time window between interactions to 30 minutes and capped the cascade duration at 2 hours to ensure causal relevance. The resulting interaction trees are collected and each node of all paths in the interaction trees are concatenated to this dataframe.

Keys:

• bee_id: Numeric unique identifier per individual bee which is a node in the tree.

• datetime: Date in year-month-day hour:minute:seconds+ms:ns format when the interaction takes place.

• x_pos: Numeric x-position of bee relative to hive at interaction start.

• y_pos: Numeric y-position of bee relative to hive at interaction start.

• vel_change_parent: Absolute velocity change of bee of parent node.

• age: Age in days of node bee. Can be NaN if the bee has no associated death date.

• is_root: Bool indicating if node is root of tree.

• depth: Depth of node in tree. E.g. depth of root is 0.

• is_leaf: Bool indicating if node is leaf of tree.

• n_children: Number of children of the subtree of the node.

• parent: Numeric bee_id of parent node.

• tree_id: Numeric unique identifier of tree.

• time_gap: Python datetime.timedelta object of time delta in between the parent and child node interaction.

• path_id: Numeric unique identifier of path.

Cosinor fit parameters - see more detailed in Cosinor data frame:

•  r_squared: R² value of the cosinor fit per bee of the day of the interaction.

•  phase: Phase of the cosinor fit per bee of the day of the interaction.

Software used to acquire and analyze the data:

speedtransfer: Cosinor fit and interaction calculation and further analyses.

bb_rhythm: Cosinor fit and interaction calculation and further analyses.

bb_behavior: Database interaction and data (pre)processing, velocity calculation.

bb_utils: Database settings and interaction.

slurmhelper: A package for slurm script handling.