ABACBS EMCR Hackathon: Spatial Biology

4 November 2024

0. In this repository

In this repository you will find R Markdown scripts for processing the raw data for the EMCR Hackathon. You do not need to re-run these scripts, but we include enough information for you to able to do so if needed.

The Hackathon Data Bundle can be downloaded in this link, and correspond to the files within the ../analysisOutput/ and ../processedData/ folders. Please download the processed data, unzip and take some time exploring the objects.

It is also useful to familiarise with the MoleculeExperiment and SpatialExperiment packages on Bioconductor.

1. Overarching task

What is given:

• Reference scRNA-seq dataset with cell-type annotation

• MoleculeExperiment object containing:

  • molecules
  • boundaries of cells according to scheme 1
  • boundaries of cells according to scheme 2

• Cell type annotation for spatial data according to scheme 1

Questions - Method:

• How can we best extract an annotation for spatial data according to scheme 2?
• How can we obtain a tissue-level characterisation independent of segmentation schemes?
• How do our approaches change if we do not have a reference dataset?

Questions - Evaluation:

• What metrics can be used to assess consistency, accuracy, stability?
• Can we identify cells that are most and least consistent, according to each scheme? Open for participants to use data & scenario as inspiration!

2. Installation required software

Please install the following R packages:

library(MoleculeExperiment) # for data structure
library(SingleCellExperiment) # for data structure
library(SpatialExperiment) # for data structure
library(scater) # for visualisation
library(patchwork) # for plotting

For reproducing the processed data please install the following R packages and scripts:

library(MoleculeExperiment) # for data structure
source(url("https://raw.githubusercontent.com/SydneyBioX/SpatialUtils/refs/heads/main/R/tileBoundaries.R"))
library(SingleCellExperiment) # for data structure
library(zellkonverter) # for reading scRNAseq h5ad data
library(scran) # for normalisation
library(scater) # for visualisation
library(randomForest) # for predictive model
library(SpatialExperiment) # for data structure
library(patchwork) # for plotting

3. Data and assumed folder structure

We assume that this repository is located inside the scripts folder of the hackathon project directory. We assume there are the following additional folders with relative paths:

• ../processedData/

  • xenium_pancreas_me.Rds - MoleculeExperiment of Xenium pancreas
  • xenium_pancreas_features.Rds - character vector of features
  • sc_pancreas_sce.Rds - SingleCellExperiment of Tabula Sapiens scRNA-seq reference
  • xenium_pancreas_spe_cells.Rds - SpatialExperiment of cell-summarised Xenium data
  • xenium_pancreas_spe_tiles.Rds - SpatialExperiment of tile-summarised Xenium data

• ../analysisOutput/

  • pancreas_colours.Rds - character vector of cell type colours
  • sc_cell_type_RF_fit.Rds - Fitted Random Forest model object used for predicting cell types

• (for reproducing data processing) ../rawData/

  • 9af4585d-a1aa-4109-913f-6fe73772af70.h5ad - H5AD file of reference single cell data
  • Xenium_V1_human_Pancreas_FFPE_outs/ - folder containing raw Xenium data bundle

Contact

Please contact the Hackathon leads Dr Shila Ghazanfar and EMCR guide Dr Yue Cao first via the ABACBS Slack channel. Further questions can be sent to [email protected].