Human_DLPFC_Deconvolution

This is an integrative, multi-assay project including individual-matched data generated from human dorsolateral prefrontal cortex (DLPFC) using spatially-resolved transcriptomics with Visium (10x Genomics), single nucleus RNA-seq with Chromium (10x Genomics), bulk RNA-seq, single molecule fluorescent in situ hybridization (smFISH) with RNAScope (Advanced Cell Diagnostics) in combination with immunofluorescence (IF). RNAScope images were processed with HALO (Indica Labs). This dataset can be used to benchmark computational deconvolution algorithms for bulk RNA-seq data that use snRNA-seq reference data.

Study design

Experimental design overview and exploration of gene detection in different assays. A. Human postmortem brain dorsolateral prefrontal cortex (DLPFC) tissue blocks across the anterior to posterior axis from 10 donors were dissected for a total of 19 tissue blocks, these tissue blocks are a subset of the 30 tissue blocks that were used in a previous spatial transcriptomic study. For each block, sequential slides were cut for different assays while maintaining the same white matter vs gray matter orientation. B. snRNA-seq data, generated as part of the same spatial transcriptomic study was collected for 19 tissue blocks Huuki-Myers et al., from which bulk RNA-seq data was also generated across two library preparations (polyA in purple or RiboZeroGold in gold) and three different RNA extractions targeting different cell fractions: cytosolic (Cyto, light color), whole cell (Total, intermediate color), or nuclear (Nuc, dark color) in this study. C. tSNE plot of the reference snRNA-seq data at the broad cell type resolution. D. Scatter plot of bulk RNA-seq principal components (PCs) 1 and 2. PC1 is associated with library type and PC2 with RNA extraction method. Colors are the same as groups in B. E. Volcano plots for the differential expression analysis between polyA and RiboZeroGold, faceted by RNA extraction method. The colors of the points are the same as B. Horizontal dotted line denotes FDR < 0.05 cutoff, vertical dotted lines are logFC = -1 and 1. F. Volcano plot for the differential expression analysis between Total bulk RNA-seq (point colors same as E) and snRNA-seq (blue points). Annotations are the same as E.

Citation

We hope that this repository will be useful for your research. Please use the following BibTeX information to cite this code repository as well as the data released by this project. Thank you!

Benchmark of cellular deconvolution methods using a multi-assay reference dataset from postmortem human prefrontal cortex.

Louise A. Huuki-Myers, Kelsey D. Montgomery, Sang Ho Kwon, Sophia Cinquemani, Nicholas J. Eagles, Daianna Gonzalez-Padilla, Sean K. Maden, Joel E. Kleinman, Thomas M. Hyde, Stephanie C. Hicks, Kristen R. Maynard, Leonardo Collado-Torres.

bioRxiv 2024.02.09.579665; doi: https://doi.org/10.1101/2024.02.09.579665

@article {Huuki-Myers2024.02.09.579665,
	author = {Louise A. Huuki-Myers and Kelsey D. Montgomery and Sang Ho Kwon and Sophia Cinquemani and Nicholas J. Eagles and Daianna Gonzalez-Padilla and Sean K. Maden and Joel E. Kleinman and Thomas M. Hyde and Stephanie C. Hicks and Kristen R. Maynard and Leonardo Collado-Torres},
	title = {Benchmark of cellular deconvolution methods using a multi-assay reference dataset from postmortem human prefrontal cortex},
	year = {2024},
	doi = {10.1101/2024.02.09.579665},
	publisher = {Cold Spring Harbor Laboratory},
	journal = {bioRxiv}
}

Data access

Files for this project are publicly available.

snRNA-seq

As documented in the spatialDLPFC project, the FASTQ files are available via Globus endpoint 'jhpce#DLPFC_snRNAseq' endpoint as well as the PsychENCODE Knowledge Portal through https://doi.org/10.7303/syn51032055.1 or https://www.synapse.org/#!Synapse:syn51032055/datasets/.

bulk RNA-seq

The RNA-seq FASTQ files are available via Globus endpoint 'jhpce#humanDeconvolutionBulkRNAseq' endpoint. Bulk RNA-seq FASTQ files are also available at NIH BioProject under accession PRJNA1086804 and Sequence Read Archive study SRP494701.

smFISH data

The RNAscope images are available via the Globus endpoint 'jhpce#humanDeconvolutionRNAScope'.

These images were analyzed with HALO software (Indica labs). The HALO exported setting files and data CSV files are available at raw-data/HALO. The combined HALO output data is available into an R object is available at processed-data/03_HALO/halo_all.Rdata.

Visium and Visium SPG data

Check the spatialDLPFC project for more details on the spatially-resolved transcriptomics data that was generated from these tissue blocks.

Code structure

Files are organized following the structure from LieberInstitute/template_project. Scripts include the R session information with details about version numbers of the packages we used.

Internal

JHPCE location: /dcs04/lieber/lcolladotor/deconvolution_LIBD4030/Human_DLPFC_Deconvolution.

• snRNA-seq: available in a file called sce_DLPFC.Rdata located at the subdirectory DLPFC_snRNAseq/processed-data/sce/.
• bulk RNA-seq: located at the subdirectory Human_DLPFC_Deconvolution/processed-data/01_SPEAQeasy/.

10X Chromium single-nucleus RNA-seq datasets

The snRNA-seq data is available as a SingleCellExperiment object.

Bulk RNA-seq datasets

Bulk RNA-seq data are available as a SummarizedExperiment object.

RNAScope/HALO image datasets

Image data is generated for RNAScope slides by analysis with HALO and outputting the analysis results as .csv tables. These tables are read from the file tree located at the subdirectory raw-data/HALO/.

Note that RNAScope experiments were performed with two combinations of markers, called Circle and Star, respectively, These experiments are distinct in that they each comprise of an analysis of an independent, albeit adjacent, tissue and each includes a different set of molecular markers (see table below).

Variables and data dictionaries

The deconvolution project makes use of a number of metadata attributes and variables in the results files mentioned about. This section describes the key terms and definitions of these attributes and variables for the deconvolution method paper.

Cell type labels

Cell type labels for snRNA-seq datasets are determined from the variable cellType_broad_hc. This can be accessed from the sce object in various ways such as sce$cellType_broad_hc or sce[["cellType_broad_hc"]]. Note that the deconvolution methods paper focuses on just 6 cell types of interest, and these are identified from among the cell type labels in the cellType_broad_hc variable.

Cell type labels for RNAScope experiments are obtained from the image analysis outputs produced by the HALO software. In brief, outputs each contain a series of columns corresponding to the cell type makers. Since each row in these outputs corresponds to an individual detected nucleus, we simply look at which marker is positive for that nucleus to determine its cell type. Cell type proportions and abundances are then calculated from these outputs.

Cell type labels aren't available for the bulk RNA-seq and other datasets produced for this project.

RNAscope labels and marker types

The molecular markers for the Circle and Star RNAscope experiments are as follows:

	cellType	marker	Combo	Type	LongName
1	Endo	CLDN5	Circle	Ab	Claudin_5
2	Astro	GFAP	Circle	Ab	GFAP
3	Inhib	GAD1	Circle	RNA_probe	GAD1
4	Excit	SLC17A7	Star	RNA_probe	SLC17A7
5	Micro	TMEM119	Star	Ab	TMEM119
6	Oligo	OLIG2	Star	Ab	OLIG2

Terminology

Throughout the deconvolution project we try to use standard terms to refer to key project entities. Here are some of the key terms to be aware of when using these project files and understanding analysis outputs.

The deconvolution equation

We take deconvolution to be the prediction of cell type amounts in a mixed sample by leveraging data from a non-mixed sample. For the deconvolution method paper, we focus on predicting either cell abundances or proportions for each of 6 cell types in bulk RNA-seq data by leveraging a single-nucleus RNA-seq reference dataset. The deconvolution equation looks like:

$Y = \pi * Z$

Where $Y$ is a matrix of mixed signals having dimensions $[G,J]$, $\pi$ is a vector of cell type amounts (proportions or abundances), and $Z$ is a matrix of non-mixed signals having dimensions $[G,K]$. The dimensions of these objects include the $G$ marker gene set, $J$ samples with mixed signal, and $K$ cell types.

Strict deconvolution

Strict deconvolution refers to solving for $\pi$ given a set of matrices $[Y,Z]$. In other words, if we assume we have a set of matrices $Y$ and $Z$, we then solve for $\pi$ based on these.

Deconvolution preparation

This simply refers to steps taken to prepare $Y$ and $Z$ prior to performing deconvolution. This may include differential weighting of the marker genes, bias corrections, data transformations and rescaling, etc.

Marker genes

The $G$ marker genes represent markers of cell types which we can use in deconvolution. Note, this is the "row" dimension in the $Y$ and $Z$ matrices.

Samples

The $J$ samples are the samples having mixed signal, for which we predict cell type amounts. Note, this is the "column" dimension in the matrix $Y$.

Cell types

The $K$ cell types represent the cell types for which we deconvolute signal and obtain amount estimates. Note, this is the "column" dimension in matrix $Y$ and the length of the vector represented by $\pi$.

Funding sources

• NIMH (USA) grant R01 MH123183
• LIBD