This repository contains necessary scirpts to reproduce all the results and figures in the MIST manuscript.

Dependencies

We recommend using a conda environment to automatically install all required dependencies. Conda installation guide can be found at https://docs.conda.io/projects/conda/en/latest/user-guide/install/index.html. After installing conda, run the following command to install a ReST environment:

conda env create -f environment.yml

Or, users could manually install all required dependencies as below:

• pandas=0.25.3
• numpy=1.18.5
• matplotlib=3.3.4
• statsmodels=0.12.0
• scipy=1.6.1
• tqdm=4.56.0
• imageio
• alphashape
• descartes
• joblib
• gseapy
• python=3.9

Reproduce figures in the manuscript

All the code to reproduce the latest figures and quantification results are in folder october22. Folder submission_ini corresponds to the first version of the code and august2022 includes code for the paper revision in August 2022. Title of each jupyter notebook indicates which figures it generates.

Data for manuscripts

Data sets are too large to be shared. All data sets are publicly available with links specified in the supplementary table 1 in the manuscript.

A sample melanoma data is shared for a demonstration at MIST github repo: https://github.com/linhuawang/MIST.

Raw count data (without imputation):

Raw count data for 4 relatively small datasets are provided at our previous repo: https://github.com/LiuzLab/MIST_manuscript.

• ./data/MouseWT/raw.csv
• ./data/MouseAD/raw.csv
• ./data/Melanoma/raw.csv
• ./data/Prostate/raw.csv

The other nine data sets used in the manuscript could be downloaded from 10X.

External validation data

Another single-cell RNA-sequencing cohort, which includes 16 mice brains with expressiondata for 37089 single cells published by Methodios Ximerakis in 2019 (GSE129788)was used to further validate the co-expression patterns of the two pairs of genes, Cldn11-Arhgef10and Gfap-Aqp4.

Ximerakis, M. et al.Single-cell transcriptomic profiling of the aging mouse brain. Nat. Neurosci.22, 1696–1708 (2019).

Holdout experiments

Run experiments

Resulted data will be available by running scripts/run_holdout_experiments.sh. This bash scripts will impute all the 13 datasets using all benchmarked methods.

Evaluation

Run evalHO.sh to generate evaluation results for further analysis and visualization.

Clustering benchmarking

Notebook analysis_scripts/2-Figure2f.ipynb was used to generate benchmarking results from different clustering methods to compare region detection accuracy. This script was also used to evaluate the performance. R script analysis_scripts/runBayesSpace.R was used to BayesSpace with different parameters (2 modes).

Figures for manuscript

1. Figure 2 and extended (run in the listed order):

• 1-Fig2a-e.ipynb
• 2-Figure2f.ipynb

2. Figure 3 and extended:

• 3-Figure3a-c-holdout-performance.ipynb
• 4-Figure3d-h.ipynb
• 5-Figure3i-and-supplementary.ipynb

3. Figure 4 and extended:

• 6-Figure4a-c&supp.ipynb
• 7-Figure4d-f.ipynb

4. Figure 5 and extended:

• 8-Figure5a&h-ABA.ipynb
• 9-Figure5b-g&i-n.ipynb

5. Other extended and supplementary figures:

• 11-Ext-Fig 7-augmentation-improvement.ipynb

Note to authors

The path to the data and code on server is at: /houston_20t/alexw/ST/MIST_additional_holdout.