Cell composition inference and identification of layer-specific transcriptional profiles with POLARIS

This repository contains the source code of POLARIS and all the code, preprocessed data used in the analysis. This repository is currently under construction.

POLARIS main functionalities

POLARIS is a versatile and generally applicable method for the analysis of spatial transcriptomics (ST) data. In particular, POLARIS has the following three main functionalities:

estimate cell type composition of each spatial spot;
detect layer-specific differentially expressed (LDE) genes, where layers refer to different anatomical or functional regions;
infer layer sub-structures solely from histological images.

POLARIS usage

Please begin by cloning this repository. After cloning the repository, the following commands can be used to perform deconvolution. In addition, please download mae_visualize_vit_large_ganloss.pth from https://github.com/facebookresearch/mae and put it in the same folder as the python scripts. POLARIS will also output the layer-specific parameters. Here is an example of performing deconvolution on the developing human heart data with and without image.

no image

username:~$ git clone https://github.com/JiawenChenn/POLARIS
username:~$ cd ./POLARIS/source_code
username:~$ # replace the following path by your own path
username:~$ python ./train.py \
--sc_cnt ../data/heart/ISS/ref/development_heart.scRNA.processed.cnt.genexrow.tsv  \
--sc_labels ../data/heart/ISS/ref/development_heart.scRNA.processed.mta.tsv  \
--sc_transpose  \
--st_cnt ../data/development_heart/PCW6.5_1/PCW6.5_1_st_cnt.tsv  \
--st_source simluation  \
--st_label ../data/development_heart/PCW6.5_1/heart_4layer.tsv  \
--st_batch_size 1024  \
--sc_batch_size 1024  \
--st_epochs 20000 \
--sc_epochs1 20000 \
--gpu \
--out_dir ./heart_noimage/ \
--prefix heart_noimage

with image

username:~$ python ./train.py \
--sc_cnt ../data/heart/ISS/ref/development_heart.scRNA.processed.cnt.genexrow.tsv  \
--sc_labels ../data/heart/ISS/ref/development_heart.scRNA.processed.mta.tsv  \
--sc_transpose  \
--st_cnt ../data/development_heart/PCW6.5_1/PCW6.5_1_st_cnt.tsv  \
--st_source human_heart  \
--st_label ../data/development_heart/PCW6.5_1/heart_4layer.tsv  \
--st_batch_size 1024  \
--sc_batch_size 1024  \
--st_epochs 20000 \
--st_epochs2 20000 \
--sc_epochs1 20000 \
--use_image \
-lr 0.001 \
--gpu \
--out_dir ./heart_image/ \
--prefix heart_image

Detection of layer-specific differentially expressed (LDE) genes

The codes used for calculate log2 fold change for the simulation results are in the down_stream folder.

Citation

Jiawen Chen et al. ,Cell composition inference and identification of layer-specific spatial transcriptional profiles with POLARIS.Sci. Adv.9,eadd9818(2023).DOI:10.1126/sciadv.add9818