nf-core/pixelator

Edit

Introduction

This document describes the output produced by the pipeline.

The pipeline is built using Nextflow and processes data using multiple subcommands of pixelator.

The directories listed below will be created in the results directory after the pipeline has finished. All paths are relative to the top-level results directory.

Proximity Network Assay (PNA)

Pipeline overview

The PNA pipeline consists of the following steps:

• Edgelist
• Aggregate
• Graph construction
• Graph collapse
• Sample Calling
• Denoising
• Analysis
• Layout creation
• Per run report generation
• Pipeline information

The output of the Proximity Network Assay (PNA) pipeline is organized into several directories, each corresponding to a specific step in the pipeline. Below is an overview of the output structure and the files generated at each step:

Edgelist

The edgelist step uses pixelator single-cell-pna edgelist to build a per-read edgelist directly from the raw FASTQ files, for both single-end and paired-end data. Reads are matched to the markers of the panel, and reads are dropped based on their quality, UMI complexity and similarity between the UMI and the pixel binding sequences.

Each FASTQ input (e.g. each lane of a sample) is processed independently. Output files are named after the FASTQ file, without the extension and without the R1/R2 suffix for paired-end data.

Aggregate

This step uses the pixelator single-cell-pna aggregate command.

The aggregate command merges the edgelists of all the FASTQ inputs of a sample (or pool) and collapses them into unique molecules, computing a read count for each molecule.

The output of this step is used as an edge list input for the graph construction step.

Graph construction

This step uses the pixelator single-cell-pna graph command. The input is the edge list parquet file generated in the aggregate step. The graph step will attempt to resolve any multiplet due to erronous edges in the graph, it will then find the connected components of the graph (i.e. the putative cells) and assign a unique ID to each component.

From this step and onwards, the output file are in PXL format. This is a custom format used by pixelator to make PNA data easier to work with. Internally it used duckdb to store the data. For more information on the PXL format, please refer to the pixelator documentation.

Graph collapse

This step uses the pixelator single-cell-pna graph-collapse command. It collapses likely UMI sequencing errors in each component of the graph. Since a UMI read error is unlikely to recur, an erroneous UMI typically shows up as a leaf node in the graph. Each such node is renamed to the nearest node of the same role whose UMI is within a given Hamming distance, and the corrected graph is saved as a new PXL file.

This step can be skipped with --skip_graph_collapse.

Denoising

This step uses the pixelator single-cell-pna denoise command. It will try to find differences between the well-connected parts of each component graph and the less well-connected parts of the graph. It will then try to find differences in marker profiles between these two parts of the graph and use these differences to denoise the graph. This reduces the effect of marker bleed-over due to incorrect cutting in the graph step.

The denoised graph will be saved as a new PXL file.

Sample Calling

This step uses the pixelator single-cell-pna sample-calling command. It takes the graph PXL file and the sample sheet as input and assigns each component to the most likely sample in pooled experiments.

Components with sample enrichment below pna_sample_calling_enrichment_threshold (default 10.0) are marked as undetermined and can be saved to a separate output file. The resulting per-sample dehashed PXL files are then used as input for the analysis step.

Analysis

This step uses the pixelator single-cell-pna analysis command to calculate spatial statistics.

Compute layouts for visualization

This step uses the pixelator single-cell-pna layout command.

It will generate precomputed layouts that can be used to visualize cells as part of the downstream analysis. This data will be appended to a PXL file.

This entire step can also be skipped using the --skip_layout option.

Generate report for all samples

This step uses pixelator-es to generate a Proxime Experiment Summary that contains information about all the samples in the run.

It will collect metrics and outputs generated by previous stages and generate a report in HTML format.

More information on the report can be found here.

The output from this step will be placed in the output folder root.

Pipeline information

Nextflow provides excellent functionality for generating various reports relevant to the running and execution of the pipeline. This will allow you to troubleshoot errors with the running of the pipeline, and also provide you with other information such as launch commands, run times and resource usage.

Output directory structure

With default parameters, the pixelator pipeline output directory will only include the latest PXL file generated by the pipeline (with the most “complete” information) and an interactive HTML report summarizing the data for all samples. The PXL dataset files can be from either the graph, analysis or layout step.

With default parameters, the <sample-id>.layout.pxl will be copied to the output directory.

Various flags are available to store intermediate files and are described in the input parameter documentation. Alternatively, you can keep all intermediate files using --save_all.

Below is an example output structure for a pipeline run using the default settings.

• pipeline_info/
• pixelator/

  • logs/

    • <sample-id>/:
      • *.log

  • pbmcs_unstimulated.layout.pxl

  • pbmcs_unstimulated.qc-report.html

  • uropod_control.layout.pxl

  • uropod_control.qc-report.html