RNASeq workflow

This repository will eventually contain the standardized analysis protocol used in the Resendis lab. At the moment this is still in planning and benchmarking phase.

Steps and tool candidates

Visualization and QC

• IGV -> https://www.broadinstitute.org/igv/
• FastQC -> www.bioinformatics.babraham.ac.uk/projects/fastqc/

Read mapping/alignment

• STAR -> https://github.com/alexdobin/STAR
• kallisto -> https://pachterlab.github.io/kallisto/
• TopHat -> https://ccb.jhu.edu/software/tophat/index.shtml
• salmon -> http://combine-lab.github.io/salmon/
• HTSeq -> http://www-huber.embl.de/HTSeq/
• RSubread -> https://bioconductor.org/packages/release/bioc/html/Rsubread.html
• StringTie -> http://www.nature.com/nbt/journal/v33/n3/full/nbt.3122.html
• RSEM -> https://github.com/deweylab/RSEM/ (used by TCGA)

Normalization and Differential Expression

• DeSeq2 -> https://bioconductor.org/packages/release/bioc/html/DESeq2.html
• edgeR -> https://www.bioconductor.org/packages/3.3/bioc/html/edgeR.html
• sleuth (for kallisto and salmon) -> https://pachterlab.github.io/sleuth
• limma + voom -> https://www.bioconductor.org/packages/release/bioc/html/limma.html

Benchmarks

• DE gene calling -> http://www.genomebiology.com/2013/14/9/R95 limma voom wins, edgeR and DeSeq also good, cuffdiff very high FPR
• DE Gene calling (high replicates) -> http://arxiv.org/pdf/1505.02017v2.pdf edgeR, DeSeq and limma the best but very conservative
• Counting -> http://arxiv.org/pdf/1505.02710.pdf (kallisto paper) kallisto best, RSEM 2nd, Tophat + Cufflinks worst
• Isoform quantification -> http://mikelove.github.io/eurobioc2015 In accuracy alpine (not readu yet) best, kallisto and RSEM 2nd, cufflinks the worst

Workflow options

• Bowtie2 + Tophat2 + Cufflinks
• STAR + RSEM
• STAR/Tophat2 + HTSeq
• Kallisto
• Salmon