/SD2

Primary LanguagePython

SD2: Spatially resolved transcriptomics deconvolution through integration of spatial and dropout information


Pipeline

Training process


Extract the dropout genes and generate pseudo-ST spots The input files are following:

  1. sc_matrix (The expression profile of scRNA-seq data whose rows are genes and columns are cells)
  2. st_matrix (The expression profile of ST data whose rows are genes and columns are spots)
  3. cell_type (The cell-type annotation vector of scRNA-seq data whose value is cell-type name corresponding with each cell)
  4. st_location (The data frame of coordinates for ST spots whose rows are spots indices and columns are "X" and "Y" representing the X and Y coordinates of each spot)
source(SD2_utiles.R)

SD2(sc_matrix,
    st_matrix,
    cell_type,
    ST_location = st_location,
    spot_num = 300, 
    lower_cellnum = 10,
    upper_cellnum = 20)

The first four items of process_data function are required. 'spot_num' is the number of generated pseudo-ST spots. ‘lower_cellnum' and 'upper_cellnum' indicate the range of cell number of each spot you want.

Construct the graph according to the transcriptional similarity and spatial connection and train by GCN

system(python train.py)

The final result would appear in the folder 'SD2_results' and entitled as "predict_output.csv'.

Here is an example of apply SD2 to the PDAC datasets (generated from Spatial Transcriptomics technique)

### load the SD2 utiles and the orginal PDAC data with its adjacent scRNA-seq data

source(SD2_utiles.R)
ST_data = readRDS('PDAC_GSM4100721.rds')
sc_count = ST_data$sc_count
st_count = ST_data$st_count
cell_type = ST_data$cell_type
st_location = ST_data$st_location

### apply the SD2 to PDAC

SD2(as.matrix(sc_count),
    as.matrix(st_count),
    cell_type,
    ST_location = st_location[,c('x','y')],
    spot_num = 300, 
    lower_cellnum = 10,
    upper_cellnum = 20)

system(python train.py)

Citation

Haoyang Li, Hanmin Li, Juexiao Zhou, Xin Gao, SD2: spatially resolved transcriptomics deconvolution through integration of dropout and spatial information, Bioinformatics, 2022;, btac605, https://doi.org/10.1093/bioinformatics/btac605