Dataset DOI: 10.5061/dryad.8cz8w9h3d

Description of the data and file structure

This dataset includes single-cell NFkB time trajectories processed from live-cell live-cell image tracking of RelA in BMDMs, the correpsonding mechanistic model fitted trajectories, and generated/sampled new single-cell trajectories, as described in Guo et al. (2025) "Modeling single-cell heterogeneity in signaling dynamics of macrophages reveals principles of information transmission". The live-cell live-cell image tracking of RelA in BMDMs were generated following the protocols of Adelaja, Taylor et al., (2021).

Corresponding author information

Name: Alexander Hoffmann
ORCID: https://orcid.org/0000-0002-5607-3845 
Affiliation: Institute for Quantitative and Computational Biosciences & Department of Microbiology, Immunology, and Molecular Genetics, University of California Los Angeles, Los Angeles, USA 
email: ahoffmann@ucla.edu

Alternative contact information

Name: Xiaolu Guo
ORCID: https://orcid.org/0000-0002-5740-2428
Affiliation: Institute for Quantitative and Computational Biosciences & Department of Microbiology, Immunology, and Molecular Genetics, University of California Los Angeles, Los Angeles, USA
email: gxll@pku.edu.cn

Related publication

Xiaolu Guo, Adewunmi Adelaja, Apeksha Singh, Roy Wollman, Alexander Hoffmann (ACCEPTED; Feb-2025) Modeling single-cell heterogeneity in signaling dynamics of macrophages reveals principles of information transmission. Nature Communication

Funding information

The work is supported by National Institutes of Health (R01AI173214) to A.H. A.S. acknowledges support from the National Institutes of Health (NIH), National Institute of General Medical Sciences training grants T32GM008185 and T32GM008042.

Files and variables

File: singaling_codon_para_for_python.zip

Description: The file folder for data files used for training regression model, inputs 'X' is the parameters, outputs 'y' is the signaling codons. (For supplementary notes figures)

Naming convention:

When applicable, file names follow the pattern

(01)_(02)_(03)_(04).csv, where:

• (01) denotes the input and output dataset and can be defined as:
  • X: input files, parameter values as the inputs for training the machine learning classifier
  • y: output files, signaling codons as the output for training the machine learning classifier
• (02) denotes the dataset type and can be defined as:
  • blank: data generated from the simulation of NFkB trajectories using the mechanistic model parameters inferred from experimental data.
  • neg_control_unidistr: synthetic data for negative control generated from the uniform distribution of parameters, with the corresponding signaling codons.
• (03) denotes the dataset type and can be defined as:
  • Duration: the file are for training machine learning classifier on predicting Duration signaling codon using the inferred paramters from single cells.
  • OscVsNonOsc: for oscillation signaling codon.
  • Speed: for speed signaling codon.
  • TotalActivity: for total activity signaling codon.
  • EarlyVsLate: for early vs late signaling codon.
  • PeakAmplitude: for speed signaling codon.
• (04) denotes the ligand in the stimuli:
  • TNF: TNF stimulated condition.
  • LPS: LPS stimulated condition.
  • Pam3CSK4: Pam3CSK stimulated condition.
  • PolyIC: PolyIC stimulated condition.
  • CpG: CpG stimulated condition.
• (05) denotes the dosage of ligand:
  • Low: low dose ligand stimulated condition.
  • Med: medium dose ligand stimulated condition.
  • High: high dose ligand stimulated condition.

Examples:

X_Duration_Pam3CSK_Low.csv: Inputs (parameters) for predicting signaling codon "Duration" for "low" dose of "Pam3CSK" condition.

X_neg_control_unidistr_Duration_Pam3CSK_Low.csv: Negative control inputs (parameters) generated by sampling from the uniform distribution, for predicting signaling codon "Duration" for "low" dose of "Pam3CSK" condition.

File: SAEM_proj_2023.zip

Description: The file folder for parameter estimation projects to estimate parameter distribution from the various ligand stimulated trajectories. Should be run in Monolix software (https://lixoft.com/products/monolix/). The files includes the raw rescaled experimental data in monolix required format, and the outcome results of the fitted individual parameters. (for Figure 2)

monolix_projects_description.csv lists the project information.

read_me.txt explained the details of the files.

Naming convention:

When applicable, file names follow the pattern

XGESN2024(01) is the folder for each project, where:

• (01) denotes the condition index, which is explained in Simulation_recording.xlsx.

Under each folder, the corresponding .mlxtran file is the project file that can be run in Monolix R2020, to start solving the corresponding NLMEs using SAEM algorithms. (Before running the file, delete the 'results' folder. After running, all the estimation results is saved in the 'results' folder.). results file folder saves all the results that are output/results of the Monolix project, please refer to https://lixoft.com/products/monolix/ for a more detailed description of the files under this folder.

SAEM_data_(01)480min.txt is the data file, where:

• (01) denotes the ligand in the stimuli:
  • TNF: TNF stimulated condition.
  • LPS: LPS stimulated condition.
  • Pam3CSK4: Pam3CSK stimulated condition.
  • PolyIC: PolyIC stimulated condition.
  • CpG: CpG stimulated condition.

params()_params()_params()_params()_params()_params()_NFkB_cyto_init_shift.txt is the ODE model file for monolix project, where:

• () denotes the parameter index (please refer to Guo et al., 2025, Supplementary table S1-2, for more details).

File: SAEM_proj_2024_downsampling.zip

Description: The file folder for parameter estimation projects to estimate parameter distribution from the downsampled NFkB trajectories, for checking different sample size impact on results. Should be run in Monolix software. including the raw rescaled experimental data in monolix required format, and the outcome results of the fitted individual parameters. (for Figure S2)

monolix_projects_description.csv lists the project information.

read_me.txt explained the details of the files.

Naming convention:

When applicable, file names follow the pattern

XGESN2024(01) is the folder for each project, where:

• (01) denotes the condition index, which is explained in Simulation_recording.xlsx.

Under each folder, the corresponding .mlxtran file is the project file that can be run in Monolix R2020, to start solving the corresponding NLMEs using SAEM algorithms. (Before running the file, delete the 'results' folder. After running, all the estimation results is saved in the 'results' folder.). results file folder saves all the results that are output/results of the Monolix project, please refer to https://lixoft.com/products/monolix/ for a more detailed description of the files under this folder.

down_samplilng_SAEM_data_(01)480min.txt is the data file, where:

• (01) denotes the ligand in the stimuli:
  • TNF: TNF stimulated condition.
  • LPS: LPS stimulated condition.
  • Pam3CSK4: Pam3CSK stimulated condition.
  • PolyIC: PolyIC stimulated condition.
  • CpG: CpG stimulated condition.

params52n2_params99_params101_params68_params75_NFkB_cyto_init_shift.txt is the ODE model file for monolix project.

File: MI_single_ligand.zip

Description: The file folder of single-ligand stimulation, experimental, sampling, fitting data, in a format for MI calculation.

.mat files Naming convention:

When applicable, file names follow the pattern

mutual_info_format_single_ligand_(01).mat is the signaling codons for single cells saved in a specific format that's used in the mutual information calculation (codes see Software, method details refer to Guo et al. 2025), where:

• (01) denotes data types.
  • Sampling: is the data generated from a simulation based on parameters sampled from inferred parameter distributions.
  • Fitting: is the data generated from a simulation based on parameters fitted to experimental data.
  • Ade: is from the experimental data.

MAT files

.mat files include nfkb struct with three struct fields. The following struct fields are included:

• sc_metrics: struct variable saving signaling codons (struct fields: TotalActivity, Speed, PeakAmplitude, OscVsNonOsc, Duration, EarlyVsLate) for each single cell.
• id: ligand & dose (if applicable) information
• ids: all conditions.

File: MI_denoise.zip

Description: The file folder of signaling codons for denoise simulation data format for MI calculation. corresponding to Figure 4.

.mat files Naming convention:
When applicable, file names follow the pattern

mutual_info_format_alldata_sc_(01)(02)_(03)_0331.mat is the signaling codons for single cells saved in a specific format that's used in the mutual information calculation (codes see Software, method details refer to Guo et al. 2025), where:

• (01) denotes the denoised modules.
  • NFkB_var_red: The variation of parameters in NF-κB module is reduced.
  • no_noise: The variation of parameters in all modules is reduced.
  • rcpt_red: The variation of parameters in receptor modules is reduced.
  • TAK_noise: The variation of parameters in modules other than adaptor module is reduced.
  • TAKac_NFkB_red: The variation of parameters in NF-κB module and adaptor module is reduced.
  • TAKac_red: The variation of parameters in adaptor module is reduced.
  • wt: The original variation of the parameters distribution inferred from experiments.
  • IkBo: The original variation of parameters distribution inferred from experiments, with reducded NFkB regulated IkBa transcription, 0.1x fold reduction.
  • IkBop25: The original variation of parameters distribution inferred from experiments, with reduced NFkB regulated IkBa transcription, 0.25x fold reduction.
  • IkBo_rcpt_red: The variation of parameters in receptor modules is reduced, with reduced NFkB regulated IkBa transcription, 0.25x fold reduction.
  • NFkB_noise: The variation of parameters in modules other than NFkB core module is reduced.
• (02) denotes data types.
  • 1,2,...,10: the 1st, 2nd, ..., 10th simulation replicate generated by independent sampling parameters. 
• (03) denotes scaling methods.
  • min_max_rescale: all signaling codons rescaled using min–max scaling.

MAT files

.mat files include nfkb struct with three struct fields. The following struct fields are included:

• sc_metrics: struct variable saving signaling codons (struct fields: TotalActivity, Speed, PeakAmplitude, OscVsNonOsc, Duration, EarlyVsLate) for each single cell.
• id: ligand & dose (if applicable) information
• ids: all conditions.

File: Sim_unstim_fitting_alldose_r2_codon_metric.mat

Description: The simulated data of unstimulated conditions, used for the Mutual information calculation. For example, the 'mutual_info_format_codon_single_ligand_unstim_Sampling_20230614.mat' uses unstim condition data from this file.

Variables

• collect_feature_vects: struct saving the "signaling codons" for each virtual single cell. With ligand and doses specified. 
  • collect_feature_vects.Duration, collect_feature_vects.OscVsNonOsc, collect_feature_vects.Speed, collect_feature_vects.TotalActivity, collect_feature_vects.EarlyVsLate, collect_feature_vects.PeakAmplitude: save corresponding signaling codons (rescaled z-scores). Different species are saved for each single cell, including 'nucNFkB', 'TNFR', 'TLR4', 'TLR2', 'TLR3', 'TLR9', 'IKK', 'TAK1', and 'IkBamRNA'. 'nucNFkB' corresponds to the NFkB activity trajectories, which are saved in collect_feature_vects.Duration{,}(1:9:end,:). The orders of species of other struct fields for each single cell are the same.
  • collect_feature_vects.info_ligand: ligand information of the stimuli condition.
  • collect_feature_vects.info_dose_str: dose information of the stimuli condition.
  • collect_feature_vects.info_data_type: simulation indicates simulated dataset.
• data: simulation data structure. data.model_sim and data.exp are identical simulation datasets. with each cell stimulation save in the data.info_ligand and data.info_dose. 
  • data.model_sim: saves the model simulated dynamic trajectories (μM) including 'nucNFkB', 'TNFR', 'TLR4', 'TLR2', 'TLR3', 'TLR9', 'IKK', 'TAK1', and 'IkBamRNA', where 'nucNFkB' corresponds to the NFkB activity trajectories and saved in data.model_sim{,}(1:9:end,:). Each row is one specifies a dynamic in one cell, each column is a time point, with a time interval of 5 minutes.
  • data.info_ligand: ligand information of the stimuli condition.
  • data.info_dose_str: ligand information of the stimuli condition.
  • data.info_num_cells: cell number multiplying species numbers for each condition.
  • data.order: orders of the single-cell based on specific signaling codon condition (by default is the random order).
  • data.exp: same as data.model_sim.
  • data.info_dose_index: ligand information of the stimuli condition.
• metrics: the dynamic features of all cells for each stimulated conditions. Different species are saved for each single cell, including 'nucNFkB', 'TNFR', 'TLR4', 'TLR2', 'TLR3', 'TLR9', 'IKK', 'TAK1', and 'IkBamRNA'. 'nucNFkB' are the NFkB activity trajectories (μM), which are saved in metrics{,}.time_series(1:9:end,:). The orders of species of other struct fields for each single cell are the same. Inmetrics{,}.time_series, each row is one specifies a dynamic in one cell, each column is a time point, with a time interval of 5 minutes.
• all above data variables, sharing the same order of the virtual cells. i.e., the first cell in collect_feature_vects, stimulated by 10ng/mL TNF are the same cell as in data, metrics.

File: Sim8_5_signle_ligand_codon_metric_r3.mat

Description: The simulated data for single-cell stimulus response specificity analysis, for WT; the data is responses of same individual cell to five different ligands. using the sampled parameters from the non-parameterized distribution for single-ligand simulation (See method "Generating new dataset of NFκB trajectories")

Variables

• sim_data_tbl: the table of the predicted simulation dataset via sampling parameters. With parameters in each module (receptor & core) saved. Ligand and doses specified for each virtual cell. Simulated time series of NFkB activity is saved every 5 minutes. Notice that other varialbes are also saved, such as IKK, TAK1, which is specified in sim_data_tbl.species. Thus, every 9 rows documenting the identical cells, but just different specifies measurement. NFkB is saved in sim_data_tbl.trajectory(1:9:end,:). The following columns are included:
  • parameter_module: the module of the parameters.
  • parameter_reac_num, parameter_para_num: reaction index of the corresponding parameter.
  • parameter_name: the parameter name.
  • parameter_fold_change: fold changes of the parameters compared to the representative cell values.
  • parameter_value: parameter value (units are explained in Table S1-2 of Guo et al., 2025).
  • ligand: Which ligand is applied in the stimuli.
  • dose_str: the dosage applied in the stimuli.
  • dose_val: values of dosage applied in the stimuli.
  • species: specifies saved in trajectory.
  • flag: N/A
  • type: wild type or other genotypes.
  • trajectory: saves the model simulated dynamic trajectories of 'nucNFkB', 'TNFR', 'TLR4', 'TLR2', 'TLR3', 'TLR9', 'IKK', 'TAK1', and 'IkBamRNA', where 'nucNFkB' corresponds to the NFkB activity trajectories and is saved in sim_data_tbl.trajectory(1:9:end,:). Each row is one specifies dynamic in one cell, each column is a time point, with time interval 5 minutes.
• metrics, collect_feature_vects, data have the same structure as in "Sim_unstim_fitting_alldose_r2_codon_metric.mat"

File: Sim5_codon_all5dose_metric.mat

Description: the simulated single-ligand (Figure S3 and supplementary notes). Note that 15 condtions are saved in this file, but only the last 5 conditions are presented in the paper. 

Variables

• metrics, collect_feature_vects, data have the same structure as in "Sim_unstim_fitting_alldose_r2_codon_metric.mat"

File: Sim2_fitting_alldose_codon_metric.mat

Description: The simulated single-ligand stimulation data, using the sampled parameters from the non-parameterized distribution for single-ligand simulation (See method "Generating new dataset of NFκB trajectories") (Figure S3)

Variables

• metrics, collect_feature_vects, data have the same structure as in "Sim_unstim_fitting_alldose_r2_codon_metric.mat"

File: Sim3_codon_r1_metric.mat

Description: The simulated data of extended doses of single-ligand stimulation, generated via sampling parameters from the non-parameterized distribution (See method "Generating new dataset of NFκB trajectories"). (Figure 3) (Sim3_codon_r1_metric.mat, Sim3_codon_r2_metric.mat, Sim3_codon_r3_metric.mat, Sim3_codon_r4_metric.mat, Sim3_codon_r5_metric.mat, have the same structure/description but just for different ligands.)

Variables

• metrics, collect_feature_vects, data have the same structure as in "Sim_unstim_fitting_alldose_r2_codon_metric.mat"

File: Exp_data_Ade.mat

Description: raw single-ligand stimulation experimental data, before rescaled.

Variables

• dataTbl: the data table of the experiments. Time series of NFkB activity is measured every 5 minutes, with ligand and doses specified for each cell. The following columns are included:
  • time_series: time series of experimental single-cell NFkB (A.U.), measured every 5 minutes.
  • Ligand: ligand applied to stimulate the corresponding cell.
  • Dose: dosage of the ligand applied to stimulate the cell.
  • DoseUnit: the dose units.
  • Other columns are documentation details of the experiments, not used for further analysis.

File: All_ligand_codon_2023_t33_cv_filtered_TNF.mat

Description: raw rescaled single-ligand stimulation simulation (Fitted individual cell parameters) and experimental data with signaling codons (Figure 2 and S4 heatmaps) , after removing the simulated trajectories with low CV (66% of the lower-CV trajectoreis are filtered out) and the corresponding experimental trajectories for TNF condition (See supplementary notes for more detail). The parameter values are also documented in this data file. (Figure 2 & Figure S2 & Figure S3)

Variables

• metrics, collect_feature_vects, data have similar structure as in "Sim_unstim_fitting_alldose_r2_codon_metric.mat", but with the differences as described below.
• metrics: the dynamic features for NFkB activity of all cells for each stimulated conditions. metrics{1:2:end} are experimental data, and metrics{2:2:end} are simulation data. 
• collect_feature_vects: structure saving the "signaling codons" of  NFkB activity for each virtual single cell. {1:2:end} are experimental data, and {2:2:end} are simulation data. 
• data: simulation data and experimental data structure. data.model_sim saves simulation, and data.exp are save the experiments, they correspond to one identical cell. With each cell stimulation save in the data.info_ligand and data.info_dose. 
• all above data variables, sharing the same order of the cells. i.e., the first cell in collect_feature_vects, stimulated by 10ng/mL TNF are the same cell as in data, metrics.

File: All_ligand_codon_2023.mat

Description: raw rescaled single-ligand stimulation experimental data with signaling codons. (No TNF filtered)

Variables

• metrics, collect_feature_vects, data have same structure as in "All_ligand_codon_2023_t33_cv_filtered_TNF.mat".

File: Ade_all_stim_unstim_codon.mat

Description: raw rescaled (by 15 conditions presented in Figure 2) single-ligand stimulation experimental data with signaling codons with unstimulated condition (condition 16). 

Variables

• metrics, collect_feature_vects, data have same structure as in "All_ligand_codon_2023_t33_cv_filtered_TNF.mat".

File: Sim16_IkBao_matching_5_signle_ligand_codon_metric_p25x_r1.mat

Description: The simulated data for single-cell stimulus response specificity analysis, for IkBa promoter mutant; the data is responses of the same individual cell to five different ligands. using the sampled parameters from the non-parameterized distribution for single-ligand simulation (See method "Generating new dataset of NFκB trajectories")

Variables

• sim_data_tbl, metrics, collect_feature_vects, data have the same structure as in "Sim8_5_signle_ligand_codon_metric_r3.mat"

File: simulation_denoise.zip

Description: The file folder for the simulated data for denoise different modules in NFkB network, for WT and IkBa promoter mutant. Using the sampled parameters from the non-parameterized distribution for single-ligand simulation (See method "Generating new dataset of NFκB trajectories"), denoise is achieved by setting the corresponding module's varied parameters to one fixed value (See method "Modeling different genotype") (For Figure 4-5)

Naming convention:
When applicable, file names follow the pattern

Sim(01)_(02)_(03)_heterogeneity_codon_metric(04)(05).mat, where:

• (01) denotes simulation index
• (02) denotes genotypes
  • SS, IkBas, wt_IkBo: IkBa mutant, with reduced NFkB regulated IkBa transcription.
  • blank: wild type parameters.
• (03) denotes denoise modules:
  • reduce_core: reduced variation of parameters in NFkB core module.
  • reduce_rcpt: reduced variation of parameters in rthe eceptor module.
  • reduce_TAK1ac: reduced variation of parameters in adaptor module.
  • NFkB: reduced variation of parameters in all modules but NFkB core module.
  • TAK1: reduced variation of parameters in all modules but adaptor module.
  • no: reduced variation of parameters in all modules.
  • blank: original parameter distribution without reducing the variation.
• (04) denotes replicates
  • _p25x_: for IkBa mutant, the NFkB regulated IkBa transcription is reduced by 0.25x fold change.
• (04) denotes replicates
  • 1, 2, ..., 5, or r1, r2, ...: replicate 1, 2,..., 5, generated from independent sampling.

Examples:

• Sim17_reduce_core_heterogeneity_codon_metric5.mat
  • Simulated NF-κB response trajectories of wild-type macrophages under five stimulation condition, with core module parameter heterogeneity reduced, replicate 5.
• Sim22_TAK1_heterogeneity_codon_metric1.mat
  • Simulated NF-κB response trajectories of wild-type macrophages under five stimulation condition, with only TAK1 module parameter heterogeneity, replicate 1.
• Sim18_wt_IkBo_codon_metric3.mat
  • Simulated NF-κB response trajectories of IkBa mutant macrophages under five stimulation condition, with parameter heterogeneity, replicate 3.

Variables in .mat files:

• sim_data_tbl, metrics, collect_feature_vects, data have the same structure as in "Sim8_5_signle_ligand_codon_metric_r3.mat"