Source data repository for: Bayesian machine learning analysis of single-molecule fluorescence colocalization images (Ordabayev et. al.).

This repository contains Figures, Figure supplements, Supplementary files, and the scripts and data used to generate them.

Clone repository

git clone https://github.com/ordabayevy/tapqir-overleaf.git
cd tapqir-overleaf

Install Tapqir

Follow these instructions to install Tapqir.

Download data

wget https://zenodo.org/record/5949388/files/ms_files.zip
unzip ms_files.zip && rm ms_files.zip

Analysis output of experimental data is stored in experimental folder:

• DatasetA: experimental/DatasetA
• DatasetB: experimental/DatasetB
• DatasetC: experimental/DatasetC
• DatasetD: experimental/DatasetD
• DatasetA: experimental/P10DatasetA (10x10 AOIs)
• DatasetA: experimental/P6DatasetA (6x6 AOIs)

Analysis output of simulated data is stored in simulations folder:

• Supplementary File 1: simulations/lamda*
• Supplementary File 2: simulations/seed*
• Supplementary File 3: simulations/height*
• Supplementary File 4: simulations/negative*
• Supplementary File 5: simulations/kon*
• Supplementary File 6: simulations/sigma*

These folders contain following files:

• data.tpqr: AOI images
• cosmos-channel0-params.tpqr: posterior parameter distributions
• cosmos-channel0-summary.csv: summary of global parameter values
• .tapqir: sub-folder containing internal files such as last model checkpoint, configuration file, and log files.

Figure 1

Example CoSMoS experiment.

Image file: figures/cosmos_experiment/cosmos_experiment.png

Figure 2

Depiction of the cosmos probabilistic image model and model parameters.

Image file: figures/graphical_model.png

To generate panels A, B, and C in the image, run (outpus figures/graphical_model.svg vector image):

python scripts/figures/graphical_model.py

Input data:

• experimental/DatasetA

Graphical model in panel D is located at figures/graphical_model.pdf.

Figure 2–Figure supplement 1

Extended graphical representation of the cosmos generative probabilistic model.

Image file: figures/graphical_model_extended.png

Figure 2–Figure supplement 2

The prior distributions for x and y spot position parameters.

Image file: figures/graphical_model_xy.png

To generate the image file, run:

python scripts/figures/graphical_model_xy.py

Figure 3

Tapqir analysis and inferred model parameters.

Image file: figures/tapqir_analysis.png

To generate the image file, run:

python scripts/figures/tapqir_analysis.py

Input data:

• simulations/lamda0.5 (panel A)
• experimental/DatasetA (panel B)

Figure 3-Figure supplement 1

Calculated spot probabilities.

Image file: figures/tapqir_analysis_probs.png

To generate the image file, run:

python scripts/figures/tapqir_analysis_probs.py

Input data:

• simulations/lamda0.5 (panel A)
• experimental/DatasetA (panel B)

Figure 3-Figure supplement 2

Reproduction of experimental data by posterior predictive sampling.

Image file: figures/tapqir_analysis_ppc.png

To generate the image file, run:

python scripts/figures/tapqir_analysis_ppc.py

Input data:

• experimental/DatasetA (panel A)
• experimental/DatasetB (panel B)
• experimental/DatasetC (panel C)
• experimental/DatasetD (panel D)

Figure 3-Figure supplement 3

Tapqir analysis of image data simulated using a broad range of global parameters.

Image file: figures/tapqir_analysis_randomized.png

To generate the image file, run:

python scripts/figures/tapqir_analysis_randomized.py

Input data:

• simulations/seed{0-16}

Figure 3-Figure supplement 4

Effect of AOI size on analysis of experimental data.

Image file: figures/tapqir_analysis_size.png

To generate the image file, run:

python scripts/figures/tapqir_analysis_size.py

Input data:

• experimental/DatasetA (14x14 AOIs)
• experimental/P10DatasetA (10x10 AOIs)
• experimental/P6DatasetA (6x6 AOIs)

Figure 4

Tapqir performance on simulated data with different SNRs or different non-specific binding rates.

Image file: figures/tapqir_performance.png

To generate the image file, run:

python scripts/figures/tapqir_performance.py

Input data:

• simulations/height* (panels A, B, C, D)
• simulations/lamda* (panels E, F, G, H)
• simulations/negative* (panel I)

Figure 4-Figure supplement 1

False negative spot misidentifications by Tapqir and spot-picker method.

Image file: figures/tapqir_performance_fn.png

To generate the image file, run:

python scripts/figures/tapqir_performance_fn.py

Input data:

• simulations/lamda1
• simulations/spotpicker_result.mat (spot-picker analysis output)

Figure 5

Tapqir analysis of association/dissociation kinetics and thermodynamics.

Image file: figures/kinetic_analysis.png

To generate the image file, run:

python scripts/figures/kinetic_analysis.py

Input data:

• simulations/kon0.01lamda0.01
• simulations/kon0.01lamda0.15
• simulations/kon0.01lamda0.5
• simulations/kon0.01lamda1
• simulations/kon0.02lamda0.01
• simulations/kon0.02lamda0.15
• simulations/kon0.02lamda0.5
• simulations/kon0.02lamda1
• simulations/kon0.03lamda0.01
• simulations/kon0.03lamda0.15
• simulations/kon0.03lamda0.5
• simulations/kon0.03lamda1

Figure 6

Extraction of target-binder association kinetics from example experimental data.

Image file: figures/experimental_data.png

To generate the image file, run:

python scripts/figures/DatasetB_ttfb_analysis.py
python scripts/figures/experimental_data.py

Input data:

• experimental/DatsetB

Figure 6-Figure supplement 1

Additional example showing extraction of target-binder association kinetics from experimental data.

Image file: figures/experimental_data_DatasetA.png

To generate the image file, run:

python scripts/figures/DatasetA_ttfb_analysis.py
python scripts/figures/experimental_data_DatasetA.py

Input data:

• experimental/DatsetA

Figure 6-Figure supplement 2

Additional example showing extraction of target-binder association kinetics from experimental data.

Image file: figures/experimental_data_DatasetC.png

To generate the image file, run:

python scripts/figures/DatasetC_ttfb_analysis.py
python scripts/figures/experimental_data_DatasetC.py

Input data:

• experimental/DatsetC

Figure 6-Figure supplement 3

Additional example showing extraction of target-binder association kinetics from experimental data.

Image file: figures/experimental_data_DatasetD.png

To generate the image file, run:

python scripts/figures/DatasetD_ttfb_analysis.py
python scripts/figures/experimental_data_DatasetD.py

Input data:

• experimental/DatsetD

Supplementary File 1

Varying non-specific binding rate simulation parameters and corresponding fit values

Spreadsheet file: supplementary/data1.xlsx

To generate the file, run:

python scripts/supplementary/data1.py

Input data:

• simulations/lamda*

Supplementary File 2

Randomized simulation parameters and corresponding fit values

Spreadsheet file: supplementary/data2.xlsx

To generate the file, run:

python scripts/supplementary/data2.py

Input data:

• simulations/seed*

Supplementary File 3

Randomized simulation parameters and corresponding fit values

Spreadsheet file: supplementary/data3.xlsx

To generate the file, run:

python scripts/supplementary/data3.py

Input data:

• simulations/height*

Supplementary File 4

No target-specific binding and varying non-specific binding rate simulation parameters and corresponding fit values

Spreadsheet file: supplementary/data4.xlsx

To generate the file, run:

python scripts/supplementary/data4.py

Input data:

• simulations/negative*

Supplementary File 5

Kinetic simulation parameters and corresponding fit values

Spreadsheet file: supplementary/data5.xlsx

To generate the file, run:

python scripts/supplementary/data5.py

Input data:

• simulations/kon*

Supplementary File 6

Varying proximity simulation parameters and corresponding fit values

Spreadsheet file: supplementary/data6.xlsx

To generate the file, run:

python scripts/supplementary/data6.py

Input data:

• simulations/sigma*