Bioinformatics and OMICs Group Meeting REFERENCE GUIDED RNA SEQUENCING

Hi Name: David Oliver Advisor: Dr. Shtutman Research: Understanding the role of COPZ2 silencing in cancer progression using RNA-seq to identify transcriptional changes caused by the loss of COPZ2 and its encoded microRNA. Experience: Microarray analysis, multiple RNA-seq analyses including long-read (PacBio) and short-read (illumina) sequencing experiments.

Why RNA-seq What’s the question? ◦Differential Expression ◦Differential splicing Advantage over other technologies ◦Increased sensitivity ◦Increased reproducibility RNA-Seq vs Dual- and Single-Channel Microarray Data: Sensitivity Analysis for Differential Expression and Clustering. Alina Sîrbu, Gráinne Kerr, Martin Crane, Heather J. Ruskin. Published: December 10, 2012DOI: /journal.pone

Before You Start Consult a statistician Consult your sequencing core

Actually Doing RNA-seq Minimum Requirements ◦Have consulted a statistician and your sequencing core ◦Know that your question can be answered using sequencing technology and that the experimental design is appropriate.

Actually Doing RNA-seq Minimum Requirements ◦Have consulted a statistician and your sequencing core ◦Know that your question can be answered using sequencing technology and that the experimental design is appropriate. ◦> 10,000,000 reads per sample ◦Much more depth required for differential splicing ◦≥ 3 biological replicates

Actually Doing RNA-seq Minimum Requirements ◦Have consulted a statistician and your sequencing core ◦Know that your question can be answered using sequencing technology and that the experimental design is appropriate. ◦> 10,000,000 reads per sample ◦Much more depth required for differential splicing ◦≥ 3 biological replicates ◦Access to decent amount of computing power ◦Can be done on a laptop but it takes ~ 3 weeks (ask me how I know) ◦Basic knowledge of Unix system and R

Actually Doing RNA-seq Minimum Requirements ◦Have consulted a statistician and your sequencing core ◦Know that your question can be answered using sequencing technology and that the experimental design is appropriate. ◦> 10,000,000 reads per sample ◦Much more depth required for differential splicing ◦≥ 3 biological replicates ◦Access to decent amount of computing power ◦Can be done on a laptop but it takes ~ 3 weeks (ask me how I know) ◦Basic knowledge of Unix system and R ◦Or, know someone who is willing to help you.

Actually Doing RNA-seq Suggested Pipeline ◦Quality assessment: ◦FastQC ◦FastX toolkit ◦Alignment: ◦Bowtie2/Tophat2 ◦STAR ◦NovoAlign ◦Counting reads: ◦FeatureCounts ◦Gencode annotation ◦Differential expression analysis ◦edgeR ◦Manipulating sequencing files ◦Samtools, bamtools Total RNA or mRNA RNA-Seq RNA expression levels Align to genome NovoAlign BowTie2 Normalization/ Quantification edgeR Quality Filtering Raw Reads Biological System STAR fastQC Read Counting FeatureCount Gencode Target Genome

RNA-seq Walkthrough Check some quality markers Getting the target genome Build aligner-specific indexed genome Perform alignment Do some file manipulation Get the annotation file Count reads Perform normalization and quantification

RNA-seq Walkthrough Check some quality markers Getting the target genome Build aligner-specific indexed genome Perform alignment Do some file manipulation Get the annotation file Count reads Perform normalization and quantification Total RNA or mRNA RNA-Seq Quality Filtering Raw Reads Biological System fastQC

Check some quality markers FastQC ◦Basic tool for generating reports ◦Java based ◦Does not provide tools for correcting errors (FastX toolkit) ◦ Other tools ◦FASTX toolkit: For fixing some problems with datasets (adapter trimming, readthrough error correction, etc) ◦SAMstat: A tool for alignment QC

RNA-seq Walkthrough Check some quality markers Getting the target genome Build aligner-specific indexed genome Perform alignment Do some file manipulation Get the annotation file Count reads Perform normalization and quantification Total RNA or mRNA RNA-Seq Align to genome Quality Filtering Raw Reads Biological System fastQC Target Genome

Getting the target genome

RNA-seq Walkthrough Check some quality markers Getting the target genome Build aligner-specific indexed genome Perform alignment Do some file manipulation Get the annotation file Count reads Perform normalization and quantification

Build aligner-specific indexed genome This step is performed by the aligner and takes a variable amount of time depending on the type of index used and the size of the genome to be indexed.

RNA-seq Walkthrough Check some quality markers Getting the target genome Build aligner-specific indexed genome Perform alignment Do some file manipulation Get the annotation file Count reads Perform normalization and quantification Total RNA or mRNA RNA-Seq Align to genome NovoAlign Bowtie2 Quality Filtering Raw Reads Biological System STAR fastQC Target Genome

Perform alignment tophat2 -p 12 --no-coverage-search --b2-N 1 --b2-L 32 --b2-i S,1,0.5 --b2-D b2-R 25 -o $RNAwork/ $RNAwork/Indexes/hg38_index $RNAwork/sample1.fastq Reads: Input : Mapped : (90.6% of input) of these: (14.1%) have multiple alignments (436 have >20) 90.6% overall read mapping rate.

RNA-seq Walkthrough Check some quality markers Getting the target genome Build aligner-specific indexed genome Perform alignment Do some file manipulation Get the annotation file Count reads Perform normalization and quantification

Do some file manipulation Depending on which aligner you choose to use, the aligned sequences may be output as a BAM or SAM file. ◦Sequence alignment/map format (SAM) ◦Contains all the alignment information plus room for user-defined information about the alignments ◦Binary alignment/map format (BAM) ◦A binary version of the SAM file ◦Added benefit of being much smaller and quickly accessed by other software ◦Not all software can manage the conversion from BAM back to SAM To manipulate these formats i.e. sort, remove duplicates, remove unaligned sequences, use either samtools or bamtools

RNA-seq Walkthrough Check some quality markers Getting the target genome Build aligner-specific indexed genome Perform alignment Do some file manipulation Get the annotation file Count reads Perform normalization and quantification Total RNA or mRNA RNA-Seq Align to genome BowTie2 Quality Filtering Raw Reads Biological System fastQC Read Counting Gencode Target Genome

Get the annotation file Annotation files are readily available from multiple sources ◦Gencode ( ) ◦Ensembl ( ) ◦Vega ( ) ◦RefSeq ( ) These annotation sources mainly vary in the number of non-coding RNAs which have been annotated. ◦RefSeq < Gencode < Ensembl < Vega We use Gencode

RNA-seq Walkthrough Check some quality markers Getting the target genome Build aligner-specific indexed genome Perform alignment Do some file manipulation Get the annotation file Count reads Perform normalization and quantification Total RNA or mRNA RNA-Seq Align to genome BowTie2 Quality Filtering Raw Reads Biological System fastQC Read Counting FeatureCount Gencode Target Genome

Count Reads FeatureCounts ◦We used to use HTseq-Count which was quite nice but we’ve switched to FeatureCounts because it is much, much, much faster. ◦Also comes as an R package (bioc::Rsubread)

RNA-seq Walkthrough Check some quality markers Getting the target genome Build aligner-specific indexed genome Perform alignment Do some file manipulation Get the annotation file Count reads Perform normalization and quantification Total RNA or mRNA RNA-Seq Align to genome BowTie2 Quality Filtering Raw Reads Biological System fastQC Read Counting FeatureCount Gencode Target Genome RNA expression levels Normalization/ Quantification edgeR

Perform normalization and quantification EdgeR: counts <- read.table(file = "All_counts.csv”) counts <- na.omit(counts) counts <- counts[-(which(rowSums(counts) == 0)),] ### start edgeR ### group <- factor(rep(c("DU145.miR1","DU145.miR148a","DU145.miR148b","DU145.miR152"), each =3)) y <- DGEList(counts = counts, group = group)### convert count matrix to a DGEList object design <- model.matrix(~0+group) ### Experimental design keep 10); y <- y[keep,] ### Remove genes with really low counts per million y$samples$lib.size <- colSums(y$counts) ### this re-calculates the library size after removing samples with low CPM y <- calcNormFactors(y)### calculate between sample normalization y <- estimateGLMRobustDisp(y, design)### calculate within sample normalizations (sort of) fit <- glmFit(y, design)### fit the “massaged data” to a generalized linear model ### perform Likelihood Ratio Test on each contrast ### lrt.du145.mir148a <- glmLRT(fit, contrast=c(-1,1,0,0,0,0,0,0)) lrt.du145.mir148b <- glmLRT(fit, contrast=c(-1,0,1,0,0,0,0,0)) lrt.du145.mir152 <- glmLRT(fit, contrast=c(-1,0,0,1,0,0,0,0)) ### generate a user-friendly output table ### tt.du145.mir148a <- topTags(lrt.du145.mir148a, n = Inf, sort.by = "none") tt.du145.mir148b <- topTags(lrt.du145.mir148b, n = Inf, sort.by = "none") tt.du145.mir152 <- topTags(lrt.du145.mir152, n = Inf, sort.by = "none")

Expected Results

Long Read (> 1kb) RNA-seq Long read analysis is performed with essentially the same workflow. For alignment, STAR or GMAP work equally well

Questions?