SENCA: A codon substitution model to better estimate evolutionary processes

Transcription

1 SENCA: A codon substitution model to better estimate evolutionary processes Fanny Pouyet Marc Bailly-Bechet, Dominique Mouchiroud and Laurent Guéguen LBBE - UCB Lyon - UMR 5558 June 2015, Porquerolles, France.

2 Codon Usage Bias - CUB Lysine codon usage in bacteria % of codons AAA AAG 20 0 A.dehalogenans B.radiodurans A.aeolicus E.coli B.aphidicola E.coli Lysine Usage sense codons, 20 amino acids Genes Number relative frequency of AAA codon 2/16

3 CUB How to measure it? Origins of the CUB: mutational vs. selective explanations. related trna concentration (um) Objectives: Disentangling the two hypotheses Codons usage (frequency %) Modelling evolutive processes explaining the observed CUB: at the nucleotidic (N), the codons (C) and the AA layers (A). 3/16

4 CUB How to measure it? Origins of the CUB: mutational vs. selective explanations. related trna concentration (um) Objectives: Disentangling the two hypotheses Codons usage (frequency %) Modelling evolutive processes explaining the observed CUB: at the nucleotidic (N), the codons (C) and the AA layers (A). 3/16

5 SENCA: Sites Evolution of Nucleotides, Codons and Amino-acids Let s have an example Thr ALANINE Gly Ser GCT GCC GCG GCA CCT CCG PROLINE CCC CCA Glu Asp GTT GTG VALINE GTC GTA Inspired by Yang and Nielsen 2008 (FMutSel). 4/16

6 SENCA: Sites Evolution of Nucleotides, Codons and Amino-acids N layer ALANINE T C GCT GCC C GCG GCA A CCT CCG PROLINE CCC CCA T GTT GTG VALINE GTC GTA κ : transition/transversion π n :equilibrium frequency of nucleotide n [A, C, G, T ]. 4/16

7 SENCA: Sites Evolution of Nucleotides, Codons and Amino-acids C layer ALANINE GCT GCC ALA GCT GCG ALA GCC ALA GCA GCA PRO CCG CCT CCG PROLINE CCC CCA GTT GTG VALINE GTC GTA VAL GTG φ AA (i) : codon i preference intra-aa. I codons(aa) φ AA(I) = 1 4/16

8 SENCA: Sites Evolution of Nucleotides, Codons and Amino-acids A layer ALANINE GCT GCC GCG GCA PRO CCT CCG PROLINE CCC CCA VAL GTT GTG VALINE GTC GTA ω : non-synonymous/synonymous substitution rates ψ AA : preference AA. AA amino acids ψ(aa) = 1 4/16

9 SENCA: Sites Evolution of Nucleotides, Codons and Amino-acids Inspiration and structure ALANINE GCT GCC ALA GCT GCG ALA GCC ALA GCA C T GCA A PRO CCG PRO C CCT CCG PROLINE CCC CCA VAL T GTT GTG VALINE GTC GTA VAL GTG Implemented in Bio++ 4/16

10 SENCA Formulas Instantaneous evolutionary rate from codon i to j : 0 if 2 or 3 different positions, πj k κ f (x i, x j ) synonymous transition, q ij = πj k κωf (x i, x j ) non-synonymous transition, πj k f (x i, x j ) synonymous transversion, πj k ωf (x i, x j ) non-synonymous transversion. with: and: f (x i, x j ) = log( x i x j ) 1 x i x j x i = n aai φ aa (i)ψ aai 5/16

11 Data 21 pathogeneous bacteria and archea From Lassalle et al., PLoS Genet concatenates of approx. 100 genes by increasing ENC (core genome, non-recombinant). Dataset Taxon Name Nb. of strains Nb of concatenates Mean GC % brucella Brucella spp francis Francisella tularensis mycobacterium Mycobacterium tuberculosis complex burk_mal Burkholderia Pseudomallei group yersinia Yersinia pestis clamydia Clamydia trachomatis sulfo Sulfolobus spp burk_ceno Burkholderia cenocepacia complex staph Staphylococcus aureus bifido Bifidobacterium longum acineto Acinetobacter spp campylo Campylobacter jejunii clostridium Clostridium botulinum strep_pyo Streptococcus pyogenes strep_pneu Streptococcus pneumoniae salmonella Salmonella enterica listeria Listeria spp B_anthracis Bacillus antharcis/aureus group nesseiria Nesseiria meningitidis escherichia Escherichia coli helicobacter Helicobacter pylori /16

12 Implementation Non-stationnary and homogeneous run, Intra-species runs: amino-acids preferences stationnary, Optimisation by max. likelihood: 109 parameters. Validation of the model: Comparisons with AIC criterium to standard codons model YN98xF61. SENCA is better in 68 over 78 concats. Exceptions are: 2 concats. of Brucella spp. and Burkholderia peusomallei, 1 of Salmonella enterica and Yersinia pestis, all of Sulfolobus spp. 7/16

13 Implementation Non-stationnary and homogeneous run, Intra-species runs: amino-acids preferences stationnary, Optimisation by max. likelihood: 109 parameters. Validation of the model: Comparisons with AIC criterium to standard codons model YN98xF61. SENCA is better in 68 over 78 concats. Exceptions are: 2 concats. of Brucella spp. and Burkholderia peusomallei, 1 of Salmonella enterica and Yersinia pestis, all of Sulfolobus spp. 7/16

14 What about the CUB? To disentangle between the 2 hypotheses: compute the CUB by comparison with expected if uniform > ENC index. Effective Number of Codons ENC varies between 61 and 20. ENC = 61 ENC = 32 ENC = 20 Strength of CUB 8/16

15 ENC index ENC SENCA OBS clostridium campylo francis staph sulfo B_anthracis listeria strep_pyo helico acineto clamy_trach strep_pneu yersinia escherichia salmo neisseria brucella bifido_longum mycobacterium burk_ceno Burk_mal At equilibrium, frequency of codon i is proportional to: 3 f (i) πi k φ aai (i)n aai ψ aai k=1 > ENC SENCA ENC OBS Species ordered by incr. GC obs 9/16

16 Species ordered by incr. GC obs ENC index ENC N 20 SENCA OBS clostridium campylo francis staph sulfo B_anthracis listeria strep_pyo helico acineto clamy_trach strep_pneu yersinia escherichia salmo neisseria brucella bifido_longum mycobacterium burk_ceno Burk_mal Compute ENCN : 3 f (i) πi k 1 1 > ENCN uniform CUB, k=1 9/16

17 ENC index ENC N C SENCA OBS clostridium campylo francis staph sulfo B_anthracis listeria strep_pyo helico acineto clamy_trach strep_pneu yersinia escherichia salmo neisseria brucella bifido_longum mycobacterium burk_ceno Burk_mal Compute ENC C : f (i) 1 φ aai (i)n aai 1 > ENC C ENC SENCA Species ordered by incr. GC obs 9/16

18 ENC index ENC N C SENCA OBS YN98 clostridium campylo francis staph sulfo B_anthracis listeria strep_pyo helico acineto clamy_trach strep_pneu yersinia escherichia salmo neisseria brucella bifido_longum mycobacterium burk_ceno Burk_mal ENC SENCA ENC OBS < ENC YN98, Is ENC SENCA mostly driven by ENC C? Quantifying. Species ordered by incr. GC obs 9/16

19 Quantification denc SENCA denc C + denc N CUB measured as distance to uniform usage: denc = 61 ENC, High correlation (R = 0.95, p-value<10e-16, Pearson correlation test), Relative importance of C and N: denc C denc SENCA > denc N denc SENCA. C N clostridium campylo francis staph sulfo B_anthracis listeria strep_pyo helico acineto clamy_trach strep_pneu yersinia escherichia salmo neisseria brucella bifido_longum mycobacterium burk_ceno Burk_mal 10/16

20 What about genome composition? GC content 0.8 N C A SENCA OBS YN GC clostridium campylo francis staph sulfo B_anthracis listeria strep_pyo helico acineto clamy_trach strep_pneu yersinia escherichia salmo neisseria brucella bifido_longum mycobacterium burk_ceno Burk_mal GC SENCA < GC obs SENCA combined effects of the 3 layers > AT enrichment at equilibrium. 11/16

21 What about genome composition? GC3 content 0.8 N C A SENCA OBS YN GC clostridium campylo francis staph B_anthracis sulfo listeria acineto strep_pyo clamy_trach strep_pneu helico yersinia escherichia salmo neisseria brucella bifido_longum mycobacterium GC3 SENCA more biased than GC3 YN98 : in agreement with GC3 obs, Which layer has the most important impact on GC bias? burk_ceno Burk_mal 12/16

22 Quantification dgc dgc N + dgc C + dgc A dgc N + dgcc + dgca GC bias measured as dgc = 0.5 GC (distance to uniform composition). High correlation (R=0.99, p-value<10e-16, Pearson correlation test) dgc* Slope=1.04, intersect fixed to 0. 13/16

23 Quantification dgc dgc N + dgc C + dgc A A C N clostridium campylo francis staph sulfo B_anthracis listeria strep_pyo helico acineto clamy_trach strep_pneu yersinia escherichia salmo neisseria brucella bifido_longum mycobacterium N, C and A impact on GC. burk_ceno Burk_mal 14/16

24 Conclusion Methodologically: Non-stationary model Multilayered model New statistical tools: GC and ENC of layers Biologically: Bias towards AT at equilibrium Importance of N, C and A for genomic bias. CUB mostly due to C. 15/16

25 Perspectives Extension of the model: Gene expression factor : measure of CUB intensity within a genome Biological questions: Influence on ω estimation, HIV Human interaction and consequences on HIV CUB, Can we detect recombination patterns: BGC? Thank you for your attention! Acknowledgments: Laurent Guéguen, Marc Bailly-Bechet, Dominique Mouchiroud, Florent Lassalle, Vincent Daubin 16/16

26 Omega comparisons SENCA YN98 17/16

27 SENCA: Site Evolution of Nucleotides, Codons and Amino-acids Implementation and parameters SENCA with 65 parameters: πj k equil. frequency of mutational process of j k κ transition/transversion φ aa (i) codon preference. For each AA, we have: i φ aa(i) = 1 n aai degenerescence of the AA ω = dn ds mutation rate of non-synynymous dn and synonymous ds ψ aa AA preference. We have: aa ψ aa = 1 In Bio++ ( Optimisation by max. likelihood. Homogeneous, non-stationary analysis (ψ aa stat. because intra-species analyses) 18/16