RENEE

Rna sEquencing aNalysis pipElinE

An open-source, reproducible, and scalable solution for analyzing RNA-seq data. See the website for detailed information, documentation, and examples: https://ccbr.github.io/RENEE/latest/

Table of Contents

• RENEE - Rna sEquencing aNalysis pipElinE
  • Table of Contents
  • 1. Introduction
  • 2. Overview
    • 2.1 RENEE Pipeline
    • 2.2 Reference Genomes
    • 2.3 Dependencies
  • 3. Run RENEE pipeline
    • 3.1 Biowulf
    • 3.2 FRCE
  • 4. References
  • 5. Version Notes

1. Introduction

RNA-sequencing (RNA-seq) has a wide variety of applications. This popular transcriptome profiling technique can be used to quantify gene and isoform expression, detect alternative splicing events, predict gene-fusions, call variants and much more.

RENEE is a comprehensive, open-source RNA-seq pipeline that relies on technologies like Docker²⁰ and Singularity²¹... now called Apptainer to maintain the highest-level of reproducibility. The pipeline consists of a series of data processing and quality-control steps orchestrated by Snakemake¹⁹, a flexible and scalable workflow management system, to submit jobs to a cluster or cloud provider.

^{Fig 1. Run locally on a compute instance, on-premise using a cluster, or on the cloud using AWS. A user can define the method or mode of execution. The pipeline can submit jobs to a cluster using a job scheduler like SLURM, or run on AWS using Tibanna (feature coming soon!). A hybrid approach ensures the pipeline is accessible to all users. As an optional step, relevelant output files and metadata can be stored in object storage using HPC DME (NIH users) or Amazon S3 for archival purposes (coming soon!).}

2. Overview

2.1 RENEE Pipeline

A bioinformatics pipeline is more than the sum of its data processing steps. A pipeline without quality-control steps provides a myopic view of the potential sources of variation within your data (i.e., biological verses technical sources of variation). RENEE pipeline is composed of a series of quality-control and data processing steps.

The accuracy of the downstream interpretations made from transcriptomic data are highly dependent on initial sample library. Unwanted sources of technical variation, which if not accounted for properly, can influence the results. RENEE's comprehensive quality-control helps ensure your results are reliable and reproducible across experiments. In the data processing steps, RENEE quantifies gene and isoform expression and predicts gene fusions. Please note that the detection of alternative splicing events and variant calling will be incorporated in a later release.

^{Fig 2. An Overview of RENEE Pipeline. Gene and isoform counts are quantified and a series of QC-checks are performed to assess the quality of the data. This pipeline stops at the generation of a raw counts matrix and gene-fusion calling. To run the pipeline, a user must select their raw data, a reference genome, and output directory (i.e., the location where the pipeline performs the analysis). Quality-control information is summarized across all samples in a MultiQC report.}

Quality Control FastQC² is used to assess the sequencing quality. FastQC is run twice, before and after adapter trimming. It generates a set of basic statistics to identify problems that can arise during sequencing or library preparation. FastQC will summarize per base and per read QC metrics such as quality scores and GC content. It will also summarize the distribution of sequence lengths and will report the presence of adapter sequences.

Kraken2¹⁴ and FastQ Screen¹⁷ are used to screen for various sources of contamination. During the process of sample collection to library preparation, there is a risk for introducing wanted sources of DNA. FastQ Screen compares your sequencing data to a set of different reference genomes to determine if there is contamination. It allows a user to see if the composition of your library matches what you expect. Also, if there are high levels of microbial contamination, Kraken can provide an estimation of the taxonomic composition. Kraken can be used in conjunction with Krona¹⁵ to produce interactive reports.

Preseq¹ is used to estimate the complexity of a library for each samples. If the duplication rate is very high, the overall library complexity will be low. Low library complexity could signal an issue with library preparation where very little input RNA was over-amplified or the sample may be degraded.

Picard¹⁰ can be used to estimate the duplication rate, and it has another particularly useful sub-command called CollectRNAseqMetrics which reports the number and percentage of reads that align to various regions: such as coding, intronic, UTR, intergenic and ribosomal regions. This is particularly useful as you would expect a library constructed with ploy(A)-selection to have a high percentage of reads that map to coding regions. Picard CollectRNAseqMetrics will also report the uniformity of coverage across all genes, which is useful for determining whether a sample has a 3' bias (observed in ploy(A)-selection libraries containing degraded RNA).

RSeQC⁹ is another particularity useful package that is tailored for RNA-seq data. It is used to calculate the inner distance between paired-end reads and calculate TIN values for a set of canonical protein-coding transcripts. A median TIN value is calucated for each sample, which analogous to a computationally derived RIN.

MultiQC¹¹ is used to aggregate the results of each tool into a single interactive report.

Quantification Cutadapt³ is used to remove adapter sequences, perform quality trimming, and remove very short sequences that would otherwise multi-map all over the genome prior to alignment.

STAR⁴ is used to align reads to the reference genome. The RENEE pipeline runs STAR in a two-passes where splice-junctions are collected and aggregated across all samples and provided to the second-pass of STAR. In the second pass of STAR, the splice-junctions detected in the first pass are inserted into the genome indices prior to alignment.

RSEM⁵ is used to quantify gene and isoform expression. The expected counts from RSEM are merged across samples to create a two counts matrices for gene counts and isoform counts.

Arriba²² is used to predict gene-fusion events. The pre-built human and mouse reference genomes use Arriba blacklists to reduce the false-positive rate.

2.2 Reference Genomes

Pre-built reference genomes are provided on Biowulf and FRCE for a number of different annotation versions, view the list here: https://ccbr.github.io/RENEE/latest/RNA-seq/Resources/#1-reference-genomes

If you would like to use a custom reference that is not already listed above, you can prepare it with the renee build command. See docs here: https://ccbr.github.io/RENEE/latest/RNA-seq/build/

2.3 Dependencies

Requires: singularity>=3.5 snakemake>=6.0

NOTE: <ins>Biowulf users</ins>: Both, singularity and snakemake, modules are already installed and available for all Biowulf users. Please skip this step as module load ccbrpipeliner will preload singularity and snakemake.

Snakemake and singularity must be installed on the target system. Snakemake orchestrates the execution of each step in the pipeline. To guarantee reproducibility, each step relies on pre-built images from DockerHub. Snakemake pulls these docker images while converting them to singularity on the fly and saves them onto the local filesystem prior to job execution, and as so, snakemake and singularity are the only two dependencies.

<hr> <p align="center"> <a href="#renee---rna-sequencing-analysis-pipeline">Back t