Smart-Seq2 quality control pipeline

Smart-seq2 quantification and quality control pipeline, developed for single-cell service at the Earlham Institue.

Starting with the FASTQ files generated from the SmartSeq2 experiment and a sample sheet containing the appropriate metadata, transcript level counts are generated, merged and used to produce QC metrics. A QC report document is generated for each of the plates, and the whole experiment.

The pipeline does not require internet connection, but some required files have to be precomputed.

Inputs

• Demulitplexed read data (.fastq.gz)
• Sample sheet (.csv)
• Config file

Outputs

• Sample count quantifications
• Expression counts matrix (transcript and gene-level)
• Quality control metrics and plots (per plate)
• QC report (per plate and for the entire experiment)

Tools used

• kallisto 0.45.1
• nextflow 19.04
• R 3.5.2 and R packages (scater, rmarkdown, tidyverse)
• singularity 2.4.2

Running the pipeline

Pipeline is written in Nextflow, so a run is usually initiated in the following way: nextflow run scqc_nf.sh -c config_file &

Examples of a config file and sample sheet are in the repository.

Config file

Parameters with 'params.' prefix can be passed when starting the pipeline by adding them at the end of the pipeline start call, e.g. 'nextflow run scqc_nf.sh -c scqc.config --qcoutdir=my_qc_directory'. Alternatively, they can be edited in the config file.

Parameters related to the output, organism species and this pipeline specifics.

• quantificationsdir - Directory to contain qunatifications produced for each of the samples and the counts matrices

• qcoutdir' - Directory to contain the final QC report and other QC-related files

• reads - Location of sample FASTQ files.

• species - 'Hsapiens' or 'Mmusculus'

• samplesheet - .csv file containing information about sample names, wells, control status and other sample metadata.

• plate_ids - List of plate identificators (typically 4 strings), as they appear in the names of raw data samples. This is how the pipeline merges samples into plate-level matrices which are then used for plate-level QC.

• mtnamefile - In case of non-human species, .rds file for mitochondrial gene. Leave empty ('') if human. This is a vector in R, containing the list of Ensembl transcript IDs, saved as .rds (using saveRDS())

• idx - Location of the kallisto index to use. Indices are precomputed. If you want to use a new one, one can be build with 'kallisto index'.

• pattern - The format of the FASTQ endings showing how they should be grouped, as a glob pattern

General HPC parameters:

• executor - Type of HPC scheduler.

• queue - Queue used for submitting jobs.

• memory - Memory assigned to a job.

• queueSize - Maximum number of jobs the pipeline will submit at once.

For more information on configuration file options, check out nextflow documentation.

Sample sheet

Sample sheet will be unique for every run.

It is a .csv file that has to have the following columns. Additional columns are not a problem, but are not used.

• unique_sample_id_suffix - Part of the FASTQ file name that uniquely matches to one sample. This is how the pipeline connects the raw data with sample sheet information.
• well - Row/column location of the sample on the plate. Something like A01, A02, etc. These are needed for plate position plots.
• plate_id - A string corresponding to one of the plates used in the experiment.
• number_of_cells - Number of cells in a well.
• control - TRUE if the well contains a control, FALSE otherwise.
• experiment - Name of the experiment. Can be anything as long as it's the same througout the column.

Required precomputed resources

• kallisto index
• list of mitochondrial genes
• singularity images
• biomaRt annotation