ncarrut / multiome-cluster-nextflow

replace decontX with cell bender

Notes

snRNA and QC pipelines must be run first. Each library must be from a single species (i.e., no barnyard experiments).

Steps

First, update nextflow.config.

Per-library RNA decontamination (decontX) and clustering + joint clustering

This step generates the cleaned RNA count matrices to be used for clustering, and performs the clustering on a per-library and joint basis.

Generate the config:

python bin/make-clean-and-cluster-config.py /path/to/rnaseq/results /path/to/qc/results > config.json

The config should look like this:

{
    "libraries": {
        "6616-NM-1-mm10": {
            "barcodes": "/lab/work/porchard/PL6616/work/rnaseq/results/starsolo/6616-NM-1-mm10/6616-NM-1-mm10.Solo.out/GeneFull_ExonOverIntron/raw/barcodes.tsv", # list of barcodes (corresponding to the market matrix file of RNA UMI counts). Should contain ALL barcodes, not just pass QC barcodes
            "features": "/lab/work/porchard/PL6616/work/rnaseq/results/starsolo/6616-NM-1-mm10/6616-NM-1-mm10.Solo.out/GeneFull_ExonOverIntron/raw/features.tsv", # list of genes (corresponding to the market matrix file of RNA UMI counts)
            "matrix": "/lab/work/porchard/PL6616/work/rnaseq/results/starsolo/6616-NM-1-mm10/6616-NM-1-mm10.Solo.out/GeneFull_ExonOverIntron/raw/matrix.mtx", # market matrix format file of RNA UMI counts, corresponding to the barcodes and features above.
            "pass_qc_barcodes": "/lab/work/porchard/PL6616/work/qc/results/atac-doublet-detection/6616-NM-1-mm10.rna.singlets.txt", # list of pass QC barcodes (should pass QC thresholds and be singlets)
            "thresholds": {
                "max_contamination": "0.2", # nuclei with estimated ambient contamination > this value (based on decontX output) are filtered out
                "soup_max_umi": "10" # barcodes with <= this number of total UMIs in the RNA UMI count matrix are taken to represent ambient background (passed to decontX as background matrix)
            }
        }
    }
}

Edit the config if desired, then run the pipeline:

nextflow run -resume -with-trace -params-file config.json -with-report -qs 300 --results /path/to/results /path/to/Multiome-Cluster-NextFlow/clean-and-cluster.nf

You may wish to run under nohup so that the pipeline continues to run in the background and does not terminate upon logging out of the server (nohup nextflow run ... &)

Output

• soup/*: Market matrix files for each library, representing the RNA UMI soup matrix (background matrix for decontX)
• pass-qc-nuclei-counts/*: Market matrix files for each library, representing the RNA UMI matrix of only pass-QC barcodes
• initial-cluster/*: Output of Seurat clustering on RNA UMI matrix of pass-QC barcodes, prior to running decontX. This clustering is used as the clustering for decontX. You should check that this looks reasonable as it may substantially impact the quality of decontX results
• decontX/*: Output of decontX, including decontaminated matrices.
• decontX-filtered/*: DecontX decontaminated matrices, after removing nuclei with high estimated contamination levels (i.e. applying the max_contamination threshold from the config)
• merged-counts/*: Merged market matrix RNA UMI matrix (merged across all libraries; i.e. merging the matrices in decontX-filtered). A prefix is added to each barcode to distinguish between identical barcodes from different libraries.
• cluster/no-contamination-filter/*: Output of Seurat clustering, using decontX decontaminated matrices but prior to removing nuclei with high estimated contamination levels (i.e., using matrices in decontX/)
• cluster/contamination-filter/*: Output of Seurat clustering, using decontX decontaminated matrices after removing nuclei with high estimated contamination levels (i.e., using matrices in decontX-filtered/)
• cluster/joint/*: Output of joint Seurat clustering across all libraries, using decontX decontaminated matrices after removing nuclei with high estimated contamination levels (i.e., using matrices in merged-counts/)