Alignment-free RNA-seq workflow. Charlotte Soneson University of Zurich Brixen 2017

Transcription

1 Alignment-free RNA-seq workflow Charlotte Soneson University of Zurich Brixen 2017

2 The alignment-based workflow ALIGNMENT COUNTING ANALYSIS Gene A Gene B... Gene X

3 The alignment-based workflow slow unnecessary? ALIGNMENT COUNTING inaccurate? doesn t account for all uncertainty in abundance estimates ANALYSIS Gene A Gene B... Gene X

4 The alignment-based workflow slow unnecessary? ALIGNMENT COUNTING inaccurate? doesn t account for all uncertainty in abundance estimates ANALYSIS Gene A Gene B... Gene X

5 Impact of differential isoform usage on gene-level counts T1 length = L T2 length = 2L sample 1 sample 2

6 Impact of differential isoform usage on gene-level counts T1 length = L T2 length = 2L sample 1 sample 2

7 Impact of differential isoform usage on gene-level counts T1 length = L T2 length = 2L Gene sample 1 sample reads 150 reads

8 Impact of differential isoform usage on gene-level counts T1 length = L T2 length = 2L Gene sample 1 sample 2 Gene S1 S2 Count reads 150 reads

9 Impact of differential isoform usage on gene-level counts true abundance condition A condition B TPM 10 5 Lengths: isoform 1: 12'232 bp isoform 2: 1'733 bp isoform 3: 891 bp isoform 4: 1'404 bp isoform 5: 543 bp 0 condition 2 isoform 1 isoform 2 isoform 3 isoform 4 isoform 5

10 Impact of differential isoform usage on gene-level counts read count condition A condition B COUNT Lengths: isoform 1: 12'232 bp isoform 2: 1'733 bp isoform 3: 891 bp isoform 4: 1'404 bp isoform 5: 543 bp 0 condition 1 condition 2 isoform 1 isoform 2 isoform 3 isoform 4 isoform 5

11 The isoform composition affects the observed read count for a gene Differential isoform usage* can lead to false positives and false negatives in differential gene expression analyses *differences in isoform composition between groups

12 What can we do? Consider another abundance unit that better reflects the underlying abundances ( number of transcript molecules ) Include adjustment of gene counts to reflect underlying isoform composition

13 What can we do? How can we get such values? Are they any good? Consider another abundance unit that better reflects the underlying abundances ( number of transcript molecules ) Include adjustment of gene counts to reflect underlying isoform composition How could such adjustment be done?

14 What can we do? Consider another abundance unit that better reflects the underlying abundances ( number of transcript molecules ) Include adjustment of gene counts to reflect underlying isoform composition How can we get such values? How could such adjustment be done? Are they any good? We need transcript-level information!

15 Abundance units read count for transcript i c i t i = c ir `i TPM i = 10 6 RP KM i = 10 9 fragment length t i P k t k c i `i P k c k library size library size = 10 9 P t i k (t k`k) `i length of transcript i TPM X i / RP KM i TPM i = 10 6 i

16 Abundance units read count for transcript i c i t i = c ir `i TPM i = 10 6 RP KM i = 10 9 fragment length t i P k t k c i `i P k c k library size library size = 10 9 P t i k (t k`k) `i length of transcript i TPM X i / RP KM i TPM i = 10 6 i

17 Abundance units read count for transcript i c i t i = c ir `i TPM i = 10 6 fragment length RP KM i = 10 9 t i P k t k c i `i P k c k library size = 10 9 P t i `i length of transcript i TPM X i / RP KM i TPM i = 10 6 i k (t k`k)

18 Abundance units read count for transcript i c i t i = c ir `i TPM i = 10 6 fragment length RP KM i = 10 9 t i P k t k c i `i P k c k library size = 10 9 P t i `i length of transcript i TPM X i / RP KM i TPM i = 10 6 i k (t k`k)

19 Abundance units read count for transcript i c i t i = c ir `i TPM i = 10 6 fragment length RP KM i = 10 9 t i P k t k c i `i P k c k library size = 10 9 P t i `i length of transcript i TPM X i / RP KM i TPM i = 10 6 i k (t k`k)

20 Average transcript lengths Similar to correction factors for library size, but sampleand gene-specific Weighted average of transcript lengths, weighted by estimated abundances (TPMs) Average transcript length for gene g in sample s: AT L gs = X i2g is `is, X i2g is =1 `is =e ective length of isoform i (in sample s) is = relative abundance of isoform i in sample s

21 Average transcript lengths T1 length = L T2 length = 2L AT L g1 =1 L +0 2L = L AT L g2 =0 L +1 2L =2L

22 Average transcript lengths T1 length = L T2 length = 2L AT L g1 =0.75 L L =1.25L AT L g2 =0.5 L L =1.5L

23 Average transcript lengths T1 length = L T2 length = 2L AT L g1 =0.75 L L =1.25L AT L g2 =0.5 L L =1.5L weights obtained from transcript TPM estimates

24 The alignment-based workflow ALIGNMENT COUNTING ANALYSIS Gene A Gene B... Gene X

25 The alignment-free workflow MAPPING ESTIMATION ANALYSIS Gene A Gene B... Gene X

26 The mapping step Does not provide full alignment information (i.e., no exact base-by-base alignment). Rather, finds all transcripts (and positions) that a read is compatible with. Comes in various flavors: pseudoalignment (kallisto) lightweight alignment (Salmon) quasimapping (Sailfish, RapMap) Bray et al. 2016; Patro et al. 2014; Patro et al. 2017; Srivastava et al. 2016

27 Transcript-level counts

28 Transcript-level counts

29 Gene-level counts 30

30 The estimation step Input: for each read, the equivalence class of compatible transcripts Probabilistic modeling of read generation process, with transcript abundance as parameter EM algorithm Output: estimated abundance of each transcript

31 Step 1: build transcriptome index kallisto name of index $ kallisto index -i my_transcripts.idx \ my_transcripts.fasta transcriptome fasta file Salmon $ salmon index -i my_transcripts.idx \ -t my_transcripts.fasta

32 Where to find transcript fasta?

33 Step 2: quantify number of cores kallisto name of index # bootstraps $ kallisto quant -i my_transcripts.idx \ -o results/sample1 -b 30 -t 10 \ sample1_1.fastq sample1_2.fastq output folder Salmon input fastq files libtype $ salmon quant -i my_transcripts.idx -l A \ -1 sample1_1.fastq -2 sample1_2.fastq \ -p 10 -o results/sample1 \ --numbootstraps 30 --seqbias --gcbias

34 Salmon LIBTYPE argument

35 Salmon LIBTYPE argument From v 0.7: -l A (automatic determination of libtype)

36 output kallisto Salmon

37 output kallisto [abundance.tsv] Salmon [quant.sf]

38 Comparison to alignment-based workflow Salmon/kallisto are considerably faster than traditional alignment+counting -> allow bootstrapping provide more highly resolved estimates (transcripts rather than gene) - can be aggregated to gene level can use a larger fraction of the reads don t give precise alignments (for e.g. visualization in genome browser) - but avoid large alignment files

39 kallisto and Salmon gene counts overall similar SRR Salmon (log(counts + 1)) kallisto (log(counts + 1))

40 Gene-level counts mostly similar to alignment-based approach SRR Salmon (log(counts + 1)) featurecounts (log(counts + 1))

41 kallisto and Salmon can use slightly more reads Number of mapped reads 3e+07 2e+07 1e+07 0e+00 SRR SRR SRR SRR SRR SRR SRR SRR featurecounts kallisto Salmon

42 How to get the estimated values into R?

43 How to get the estimated values into R?

44 How to get the estimated values into R? TPMs counts ATL offsets

45 A word of warning Abundance estimates for lowly expressed transcripts are highly variable and should be interpreted with caution Estimated TPM True TPM Soneson, Love & Robinson, F1000 Research 2016

46 A word of warning Problematic when coverage of region defining an isoform is low Soneson, Love & Robinson, F1000 Research 2016

47 A word of warning When aggregated to the gene level, abundance estimates are less variable Estimated TPM True TPM Soneson, Love & Robinson, F1000 Research 2016

48 References Srivastava et al.: RapMap: a rapid, sensitive and accurate tool for mapping RNA-seq read to transcriptomes. Bioinformatics 32:i192-i200 (2016) - RapMap Patro et al.: Accurate, fast, and model-aware transcript expression quantification with Salmon. biorxiv dx.doi.org/ / (2015) - Salmon Bray et al.: Near-optimal probabilistic RNA-seq quantification. Nature Biotechnology 34(5): (2016) - kallisto Patro et al.: Sailfish enables alignment-free isoform quantification from RNA-seq reads using lightweight algorithms. Nature Biotechnology 32: (2014) - Sailfish Pimentel et al.: Differential analysis of RNA-Seq incorporating quantification uncertainty. biorxiv / (2016) - sleuth Wagner et al.: Measurement of mrna abundance using RNA-seq data: RPKM measure is inconsistent among samples. Theory in Biosciences 131: (2012) - TPM vs FPKM Soneson et al.: Differential analyses for RNA-seq: transcript-level estimates improve gene-level inferences. F1000Research 4:1521 (2016) - ATL offsets (tximport package) Li et al.: RNA-seq gene expression estimation with read mapping uncertainty. Bioinformatics 26(4): (2010) - TPM, RSEM