Re-analysis Mathys et al., 2019

Now published in eLife.

A re-analysis of the Single-cell transcriptomic analysis of Alzheimer’s disease using a standardised data processing and pseudobulk differential expression approach.

Standardised processing protocol

We used scFlow (v0.6.1) for all of the 24 patients with Alzheimer's disease pathology and the 24 control patients. The config file with all the parameters is available in ./scFlow_files. This approach resulted in more stringent quality control, leading to the exclusion of more, low quality cells.

Download processed data

An single-cell object (SCE) of the Mathys et al. study into Single-nucleus transcriptomic analysis and differential expression (DE) of Alzheimer’s disease data after processing with scFlow is available on synapse see: https://doi.org/10.7303/syn51758062.1.

Note this includes the processed count matrix and associated metadata. This will be needed to run the analysis below. Also note that here we match the cell types to the original paper but we have also made available a more granular cell type groupings which you can access in the metadata - the cluster_celltype column rather than the allan_celltype column.

Run analysis

The run_reanalysis_Mathys_19.R in the R folder can be used to derive the EGs from the reprocessed Mathys et al., 2019 Alzheimer's disease patient snRNA-Seq data. This script uses a custom written function to apply pseudobulk differential analysis to any single-cell dataset (see sc_cell_type_de.R).

Docker file

We also provide a docker file to create an image to rerun the analysis - removing the need of the user to install all the dependencies themselves. Simply run the following with docker installed:

docker pull neurogenomicslab/reanalysis_mathys_2019

or recreate the docker image with:

docker build -t reanalysis_mathys_2019 .

Whether you pull or recreate the image, next run it:

docker run -e PASSWORD=reanalysis --rm -p 8787:8787 reanalysis_mathys_2019

Now navigate to localhost:8787 in a web browser and log in with: username: rstudio password: reanalysis to access the docker image. Then clone the repo in the terminal of Rstudio with:

git clone https://github.com/neurogenomics/reanalysis_Mathys_2019
#install data
cd reanalysis_Mathys_2019/
wget https://figshare.com/ndownloader/files/38819949 -O ./data/sce.qs
#rerun analysis
cd R/
Rscript run_reanalysis_Mathys_19.R

Once this runs you can look in the results folder or also knit the Rmd file (Mathys_results_comparison_pb.html) to view the results.

Comparison against original findings

See results

details

Mathys_results_comparison_pb.html gives a comparison between the results reported in the original publication and our findings using pseudobulk differential expression and a standardised data processing approach. Note the authors took cells as independent replicates, a cell-level analysis, in their work and compared the consistency in directionality and rank of their differentially expressed genes (DEGs) against a poisson mixed model. We show that these DEGs are just an artefact of taking cells as independent replicates by plotting the number of DEGs found against the cell counts. There is a strong correlation for their results but not for the pseudobulk DEGs.

Replicate Random permutation analysis

Run random_perm_pseudorep_pseudobulk_analysis.R to replicate the random permutation analysis based on pseudoreplication and pseudobulk differential expression methods.

Cite

If you want further details or are using this work please see/cite our manuscript.