Lecture 8 Understanding Transcription RNA-seq analysis. Foundations of Computational Systems Biology David K. Gifford

Transcription

1 Lecture 8 Understanding Transcription RNA-seq analysis Foundations of Computational Systems Biology David K. Gifford 1

2 Lecture 8 RNA-seq Analysis RNA-seq principles How can we characterize mrna isoform expression using high-throughput sequencing? Differential expression and PCA What genes are differentially expressed, and how can we characterize expressed genes? Single cell RNA-seq What are the benefits and challenges of working with single cells for RNA-seq? 2

3 RNA-Seq characterizes RNA molecules export to cytoplasm nucleus A B C A C mrna High-throughput sequencing of RNAs at various stages of processing splicing A B C pre-mrna or ncrna transcription A B C Gene in genome Courtesy of Cole Trapnell. Used with permission. cytoplasm Slide courtesy Cole Trapnel 3

4 Pervasive tissue-specific regulation of alternative mrna isoforms. 4 Courtesy of Macmillan Publishers Limited. Used with permission. Source: Wang, Eric T., Rickard Sandberg, et al. "Alternative Isoform Regulation in Human Tissue Transcriptomes." Nature 456, no (2008): ET Wang et al. Nature 000, 1-7 (2008) doi: / nature07509

5 RNA-Seq: millions of short reads from fragmented mrna Extract RNA from cells/tissue + splice junctions! Courtesy of Macmillan Publishers Limited. Used with permission. Source: Pepke, Shirley, Barbara Wold, et al. "Computation for ChIP-seq and RNA-seq Studies." Nature Methods 6 (2009): S Pepke et. al. Nature Methods

6 Mapping RNA-seq reads to a reference genome reveals expression Sox2 6

7 RNA-seq reads map to exons and across exons Reads over exons Smug1 Junction reads (split between exons) 7

8 Two major approaches to RNA-seq analysis 1. Assemble reads into transcripts. Typical issues with coverage and correctness. 2. Map reads to reference genome and identify isoforms using constraints Goal is to quantify isoforms and determine significance of differential expression Common RNA-seq expression metrics are Reads per killobase per million reads (RPKM) or Fragments per killobase per million (FPKM) Short sequencing reads, randomly sampled from a transcript exon 1 exon 2 exon 3 8

9 Aligned reads reveal isoform possibilities A B C identify candidate exons via genomic mapping A B A C B C Generate possible pairings of exons A B A C B C Align reads to possible junctions Courtesy of Cole Trapnell. Used with permission. Slide courtesy Cole Trapnell 9

10 We can use mapped reads to learn the isoform mixture ψ" A D C B Isoform Fraction T 1 ψ 1" T 2 ψ 2" T 3 ψ 3" T 4 ψ 4" E Courtesy of Cole Trapnell. Used with permission. Slide courtesy Cole Trapnell 10

11 Detecting alternative splicing from mrna-seq data Isoforms Inclusion reads Common reads Common reads Exclusion reads Given a set of reads, estimate: = Distribution of isoforms 11

12 P(R i T=T j ) Excluded reads If a single ended read or read pair R i is structurally incompatible with transcript T j, then P(R = R i T = T j ) = 0 R i T j Intron in T j Courtesy of Cole Trapnell. Used with permission. Slide courtesy Cole Trapnell 12

13 P(R i T=T j ) Single end reads Cufflinks assumes that fragmentation is roughly uniform. The probability of observing a fragment starting at a specific position S i in a transcript of length l j is:" P(S = S i T = T j ) = 1 l j starting position in transcript, S i R i T j Transcript length l j Courtesy of Cole Trapnell. Used with permission. Slide courtesy Cole Trapnell 13

14 P(R i T=T j ) Paired end reads Assume our library fragments have a length distribution described by a probability density F. Thus, the probability of observing a particular paired alignment to a transcript:" P(R = R i T = T j ) = F(l j(r j )) l j Implied fragment length l j (R i ) R i T j Courtesy of Cole Trapnell. Used with permission. Slide courtesy Cole Trapnell 14

15 Estimating Isoform Expression Find expression abundances ψ 1,,ψ n for a set of isoforms T 1,,T n Observations are the set of reads R 1,,R m m P(R Ψ) = Ψ j P(R = R i T = T j ) i=0 j=0 L(Ψ R) P(R Ψ)P(Ψ) n Ψ = argmaxl(ψ R) Ψ Can estimate mrna expression of each isoform number of reads that map to a gene and ψ using total 15

16 Case study: myogenesis Transcript categories, by coverage Transcripts (%) match contained intra intron novel isoform repeat other Cufflinks identified 116,839 distinct transcribed fragments (transfrags) Nearly 70% of the reads in 14,241 matching transcripts Tracked 8,134 transfrags across all time points, 5,845 complete matches to UCSC/ Ensembl/VEGA Transcripts Tracked 643 new isoforms of known genes across all points Reads per bp Courtesy of Cole Trapnell. Used with permission. Slide courtesy Cole Trapnell 16

17 Case study: myogenesis Transcript categories, by coverage Transcripts (%) match contained intra intron novel isoform repeat other ~25% of transcripts have light sequence coverage, and are fragments of full transcripts Intronic reads, repeats, and other artifacts are numerous, but account for less than 5% of the assembled reads. Transcripts Reads per bp Courtesy of Cole Trapnell. Used with permission. Slide courtesy Cole Trapnell 17

18 Lecture 8 RNA-seq Analysis RNA-seq principles How can we characterize mrna isoform expression using high-throughput sequencing? Differential expression and PCA What genes are differentially expressed, and how can we characterize expressed genes? Single cell RNA-seq What are the benefits and challenges of working with single cells for RNA-seq? 18

19 19

20 20

21 21

22 22

23 23

24 Scaling RNA-seq data (DESeq) i gene or isoform j sample (experiment) m number of samples K ij number of counts for isoform i in experiment j s j sampling depth for experiment j (scale factor) s j = median i K ij ( v=1kiv m ) 1 m 24

25 Model for RNA-seq data (DESeq) i gene or isoform p condition j sample (experiment) p(j) condition of sample j m number of samples K ij number of counts for isoform i in experiment j q ip Average scaled expression for gene i condition p q ip = 1 # of replicates j in replicates K ij s j µ ij = q ip( j) s j σ ij 2 2 = µ + s j ij ( q ) v p ip( j) K ij ~ NB ( µ ij, σ ij 2 ) 25

26 2 2 σ ij = µ ij + s jvp(q ) ip( j) Orange Line DESeq Dashed Orange edger Purple - Poission Courtesy of the authors. License: CC-BY. Source: Anders, Simon, and Wolfgang Huber. "Differential Expression Analysis for Sequence Count Data." Genome Biology 11, no. 10 (2010): R

27 Significance of differential expression using test statistics Hypothesis H0 (null) Condition A and B identically express isoform i with random noise added Hypothesis H1 Condition A and B differentially express isoform Degrees of freedom (dof) is the number of free parameters in H1 minus the number of free parameters in H0; in this case degrees of freedom is 4 2 = 2 (H1 has an extra mean and variance). Likelihood ratio test defines a test statistic that follows the Chi Squared distribution T i = 2 log P( K ia H1)P( ib P ia ib K H1) K, K H 0 ( ) P(H 0) 1 ChiSquaredCDF (T i dof ) 27

28 28 Courtesy of the authors. License: CC-BY. Source: Anders, Simon, and Wolfgang Huber. "Differential Expression Analysis for Sequence Count Data." Genome Biology 11, no. 10 (2010): R106.

29 Hypergeometric test for overlap significance N total # of genes 1000 n1 - # of genes in set A 20 n2 - # of genes in set B 30 k - # of genes in both A and B 3 P( k) =! # " n1 k $! N n1 &# %" n2 k! N $ # & " n2 % $ & % ( ) = P(i) P x k min(n1,n2) i=k

30 30

31 31

32 32

33 33

34 34

35 35

36 Lecture 8 RNA-seq Analysis RNA-seq principles How can we characterize mrna isoform expression using high-throughput sequencing? Differential expression and PCA What genes are differentially expressed, and how can we characterize expressed genes? Single cell RNA-seq What are the benefits and challenges of working with single cells for RNA-seq? 36

37 37 Courtesy of Fluidigm Corporation. Used with permission.

38 Single-cell RNA-Seq of LPS-stimulated bone-marrow-derived dendritic cells reveals extensive transcriptome heterogeneity. AK Shalek et al. Nature 000, 1-5 (2012) doi: /nature Courtesy of Macmillan Publishers Limited. Used with permission. Source: Shalek, Alex K., Rahul Satija, et al. "Single-cell Transcriptomics Reveals Bimodality in Expression and Splicing in Immune Cells." Nature (2013).

39 Analysis of co-variation in single-cell mrna expression levels reveals distinct maturity states and an antiviral cell circuit. AK Shalek et al. Nature 000, 1-5 (2012) doi: /nature Courtesy of Macmillan Publishers Limited. Used with permission. Source: Shalek, Alex K., Rahul Satija, et al. "Single-cell Transcriptomics Reveals Bimodality in Expression and Splicing in Immune Cells." Nature (2013).

40 Analysis of co-variation in single-cell mrna expression levels reveals distinct maturity states and an antiviral cell circuit. AK Shalek et al. Nature 000, 1-5 (2012) doi: /nature Courtesy of Macmillan Publishers Limited. Used with permission. Source: Shalek, Alex K., Rahul Satija, et al. "Single-cell Transcriptomics Reveals Bimodality in Expression and Splicing in Immune Cells." Nature (2013).

41 RNA-seq library complexity can help qualify cells for analysis 41 Michal Grzadkowski Michal Grzadkowski. All rights reserved. This content is excluded from our Creative Commons license. For more information, see

42 RNA-seq library complexity can help qualify cells for analysis Michal Grzadkowski Michal Grzadkowski. All rights reserved. This content is excluded from our Creative Commons license. For more information, see 42

43 43 FIN

44 MIT OpenCourseWare J / J / J / 7.36J / 6.802J / 6.874J / HST.506J Foundations of Computational and Systems Biology Spring 2014 For information about citing these materials or our Terms of Use, visit: