File naming convention

To standardlize the process, we strongly suggest to start with universal file naming format. In this pipeline, we will use filename convention as: [PD|AD|HD|HC]_[subjectID]_[cellType]_<b#>_<rep#>.[R1|R2].<fq|fastq>.gz where,

• Beginning with patient type, PD="Parkinson's disease", AD="Alzhimer's disease", HD="Huntington's disease" and HC="Healthy controls", in 2-letter abbreviation;
• Following with subjectID (NOT lane# or sequencing date), which is unique per subject;
• Following with cell type or tissue short name, e.g. SNDA="Substantial nigra dopamine neuron", MCPY="Motor cortex pyramidal neuron" etc.;
• b# and rep# are optional, in format of (if any);
• batch number: b1, b2 etc.
• replication number (capitalized one for biological rep), e.g. rep1 is for technique replicate 1, Rep1 for biological replicate 1
• Use R1 or R2 to tell the two mates of pair-end sequencing data;
• Use zipped fastq;

Folder structure

For each project (e.g. PDMap), it should have a folder named with project short name and followed by the type of data, such as PDMap_RNAseq, HDPredict_smallRNA etc.

• rawfiles
  • raw fastq files from sequencing;
  • use soft links for external location or unformated file name
  • log file (readme.txt or Excel)
• filtered
  • filtered files (e.g. adaptor removal/clip)
• run_output
  • sub-folder per sample (e.g. sample1, sample2 etc.)
    • output of Tophat/Cufflinks/htseq-count runs
    • uniq subfolder for the runs for unique reads only
    • status indication file (.status*) for tracking the progress
• for_display
  • files used for display on UCSC / IGV, such as *.bam, *.bam.bai, *.bw, *.gtf etc.
  • use soft links to the output files
• results
  • result files for integrative analysis, e.g. differential analysis by combining all samples.

Pipeline requirement

1. Install programs: tophat, cufflinks, bowtie, bedtools, samtools, fastq-mcf, fastqc, and htseq-count;
2. Install Jim Kent's executable programms: http://hgdownload.cse.ucsc.edu/admin/exe/;
3. Install R and bioconductor packages: DESeq2, MatrixEQTL, SPIA etc.
4. Add path of the executable programs to the $PATH;

Pipeline structure

Each sub-module has its own status hidden file (e.g. .status.$modulename.adaptorremoval). To check if any sample not completed at any step, use script below:

$grep touch ~/pipeline/modules/_RNAseq.sh | grep -v "#" | sed 's/.*modulename\.//g' | while read i; do echo $i `ll ~/neurogen/rnaseq_PD/run_output/*/.status._RNAseq.sh.$i | wc -l`; done

Main script:

RNAseq.pipeline.sh

• Usage: RNAseq.pipeline.sh /data/neurogen/rnaseq_PD/rawfiles
• Function: Main script for submitting RNAseq data analysis jobs to high-properformance computing cluster in batch. For now, it's configured to support only HPC environment with LSF job scheduler.
• Input: absolute path of folder for the raw fastq files
• Output: Tophat/Cufflinks/htseq-count/callSNP etc. runs for each sample using both multiple and unique mappers

Modules:

_RNAseq.sh

• Usage: _RNAseq.sh HC_BN10-39_2.R1.fastq.gz HC_BN10-39_2.R2.fastq.gz
• Function: Routine steps for pair-end RNAseq data
• Input: a pair of fastq files from PE RNAsequencing
• Output: BAM/SAM (from alignment), BED/GTF (from assembly) etc.

List of sub-moduels in the core script _RNAseq.sh:

$ grep touch ~/pipeline/modules/_RNAseq.sh | grep -v "#" | sed 's/.*modulename\.//g'
adaptorremoval 
shortReadsExtract 
shortReadsExtract_stemloop
fastqc
mapping
circRNA
sam2bam
bam2stat
sam2bw
rpm_vs_coverage
cufflinks.multi.denovo
cufflinks 
htseqcount
metaexon
uniq
uniq.sam2bw
uniq.bam2stat
cufflinks.denovo
uniq.bam2annotation
uniq.normalize
uniq.cufflinks.rpkm
uniq.cuffquant.rpkm
uniq.htseqcount
uniq.metaexon
makelinks