Scikit-ribo reveals precise codon-level translational control by dissecting ribosome pausing and codon elongation

Transcription

1 Scikit-ribo reveals precise codon-level translational control by dissecting ribosome pausing and codon elongation Han Fang October 26, 2016 Biological Data Science

2 The hidden treasure in genomics Replication Transcription Translation DNA RNA Protein RNA Polymerase Ribosome

3 The hidden treasure in genomics Replication Transcription Translation DNA RNA Protein RNA Polymerase Ribosome

4 What is ribosome profiling (Riboseq)? Ingolia et al. Science. (2009) Ingolia. Nat Rev Genet. (2014)

5 What is ribosome profiling (Riboseq)? Normal translation efficieny (TE) Ingolia et al. Science. (2009) Ingolia. Nat Rev Genet. (2014)

6 What is ribosome profiling (Riboseq)? Normal translation efficieny (TE) Less efficient translation More efficient translation Ingolia et al. Science. (2009) Ingolia. Nat Rev Genet. (2014)

7 Calculate translational efficiency (TE) Less efficient translation Normal translation efficieny (TE) More efficient translation log 2 (TE) < 0 log 2 (TE)=0 log 2 (TE) > 0 TE = Riboseq rpkm RNAseq rpkm

8 Hypothesis: TE distribution could be skewed by ribosome pausing events. Ribosome footprints without bias Ribosome footprints with pausing

9 Simulated S. cerevisiae data Null distribution

10 Simulated S. cerevisiae data - TE distribution are negatively-skewed by ribosome pausing events Null distribution With ribosome pausing sites Randomly imputed ribosome pausing sites to 20% of the genes

11 Ribosome pausing sites (peaks) finding by negative binomial mixture model Y P (X i i,µ i,k i,r i )= Y j i NB(X ij µ i,r i )+(1 i )NB(X ij k i µ i,r i ), Yifei Huang for gene i at position j, where k 5 H 0 : =1 H : =1 H 1 : 6= 1 Yes Bimodal distribution, i.e. potential pausing sites Posterior probability > cutoff No Yes Peak/Pausing sites Given a gene Likelihood ratio test Bonferroni adjusted p-value < 0.05 No Unimodal distribution, i.e. no pausing Not a pausing site

12 Ribosome pausing sites (peaks) finding by negative binomial mixture model Y P (X i i,µ i,k i,r i )= Y j i NB(X ij µ i,r i )+(1 i )NB(X ij k i µ i,r i ), Yifei Huang for gene i at position j, where k 5 H 0 : =1 H : =1 H 1 : 6= 1 Yes Bimodal distribution, i.e. potential pausing sites Posterior probability > cutoff No Yes Peak/Pausing sites Given a gene Likelihood ratio test Bonferroni adjusted p-value < 0.05 No Unimodal distribution, i.e. no pausing Not a pausing site # genes # genes (rpkm > 100) # genes with pausing # ribosome pausing sites identified

13 mrna with stronger secondary structure tend to have ribosome pausing events p-value< 2*10-16 Secondary structure score No Yes PARS scores obtained from Kerteszet al. Nature (2010)

14 mrna with stronger secondary structure tend to have ribosome pausing events p-value< 2*10-16 Secondary structure score PARS scores ribosome counts PARS_score position ribosome_pausing FALSE TRUE No Yes position PARS scores obtained from Kerteszet al. Nature (2010)

15 mrna with stronger secondary structure tend to have ribosome pausing events p-value< 2*10-16 Secondary structure score PARS scores ribosome counts PARS_score position ribosome_pausing FALSE TRUE No Yes position PARS scores obtained from Kerteszet al. Nature (2010)

16 Simplifying the generalized linear model (GLM) RNA count

17 Simplifying the generalized linear model (GLM) RNA count Secondary structure count

18 Simplifying the generalized linear model (GLM) RNA count Secondary structure count Dwell time (1/speed) count Translation count

19 Simplifying the generalized linear model (GLM) RNA count Secondary structure count Dwell time (1/speed) count Translation efficiency (TE) count

20 Joint inference of protein TE and codon dwell time using GLM, while accounting for secondary structure GLM for joint inference of TE and codon dwell time: Y ij NB(mean = µ ij,dispersion= ), for gene i, position j g(µ ij )= 0 + x i m + t i + c k + sx ij {z} {z} {z} {z } mrna TE codon secondary structure where g(.) is a log link function, µ ij = E(Y ij ), x i m is mrna abundance for gene i, i t is translational e ciency for gene i, k c is dwell time for codon k, sx ij is secondary structure e ect at position j for gene i.

21 Scikit-ribo perfectly reproduced relative codon dwell time from Weinberg et al

22 Scikit-ribo perfectly reproduced relative codon dwell time from Weinberg et al Rarest codons CGA CGG CCG

23 Significant correlation between trna abundance and codon elongation rates

24 GLM estimates vs. RPKM-based estimates reveals systematic bias in typical Riboseq analysis

25 GLM estimates vs. RPKM-based estimates reveals systematic bias in typical Riboseq analysis Highly structured mrna genes The rpkm based approach overestimated TE of highly structured mrna, while the rest of the mrna were slightly under-estimated, as hypothesized.

26 Accurate TE estimation supported by proportional synthesis for heterodimeric complexes in S. cerevisiae.

27 Summary Discussed: 1) Introduced scikit-ribo for joint analysis of Riboseq and RNAseq data. 2) Identified biases in Riboseq data due to ribosome pausing. 3) Corrected biases and revealed underlying biology 4) Joint inference of codon elongation rate and protein TE 5) Revealed precise translational control at codon level Ribosome A-site position predicition A-site codon localization Ribosome pausing site calling Translation efficiency inference

28 Acknowledgments Lyon Lab Max Doerfel Yiyang Wu Jonathan Crain Jason O Rawe Schatz Lab Fritz Sedlazeck Tyler Garvin James Gurtowski Maria Nattestad Srividya Ramakrishnan Gholson Lyon Michael Schatz CSHL JHU UCSF Rutgers Stony Brook University Yifei Huang Adam Siepel Noah Dukler Rachel Green Jonathan Weissman Joshua Dunn David Weinberg Premal Shah Rob Patro