1. Getting the best out of multiple sequence alignment methods in the genomic era Cédric Notredame Comparative Bioinformatics Group Bioinformatics and Genomics Program

2. Which Tool for Which Sequence ?

3. In- SilVo Biology In Silico Biology – Making Sense of digital data In Vivo Biology – Recording data in a living Cell In SilVo Biology – Connect In-Vitro and In-Vivo In-Vivo: High-throughput recording In-Silico: High-Throughput analysis

5. Is it Possible to Compare all Types of Sequences ? Non Transcribed World – Genes/Full Genomes Lagan, TBA – Promoter Regions Meta-Aligner Motifs Finders – Nucleosome ??? Multiple Genome Aligners – Not Very Accurate – Very Fast – Deal with rearrangements

6. Multiple Genome Alignments and re-sequencing Before – Re-sequence Human Genomes – Map the Reads onto the reference genome Now – Re-sequence – Assemble – Align – Non trivial with very large datasets

7. Is it Possible to Compare all Types of Sequences ? RNA Comparison – Less Accurate than Proteins – Secondary Structures ncRNA World – Sankoff Time O(L 2n ) Space O(L 3n ) – Consan – R-Coffee

8. Is it Possible to Compare all Types of Sequences ? Protein Comparisons – Very Accurate – 3D-Structure Improves it Protein Aligners – ClustalW – T-Coffee – 3D-Coffee

9. What Changes with 1000 Genomes?

10. Phylogeny Vs Function Function – Low level => Biochemistry => Protein Domains – High Level => Metabolic Pathway => Orthology Orthology – Phylogenetic Analysis – Phylogenetic Analysis =>Accurate Alignments

11. Using The tree Correct Tree  Correct Orthologous Assignment  Correct Functional Prediction

12. The Alignment that Hides The Forest…

13. Why So Much interest for MSA methods???

14. Phylogenetic Trees and Multiple Sequence Alignments

16. Positive Selection and MSAs

19. Multiple Genome Alignments

21. Genomic Era: The Goal 10.000 Sequences: interspecies 1 Billion: Re-sequencing Incorporation of ALL experimental Data – Structure, Genomic, ChIp-Chip, ChIp-Seq… Alignments suitable for all applications of comparative genomics – Homology Modeling (function) – Functional Analysis – Phylogenetic Reconstruction – 3D-Modelling Accurate Alignments for ALL kind of data Non Transcribed DNA Transcribed DNA Translated DNA

22. Genomic Era Challenges Accuracy – Proteins: 30% is the limit – DNA/RNA 70% is the limit Scale – With too many sequences algorithms lose in accuracy Data Integration – Structure – Homology – Genomic Structure – Function – Proteomics Methods – Wealth of alternative methods – Poorly Characterized

23. Consistency and Data Integration Most methods rely on the progressive algorithm Consistency based methods have been designed as an extension Consistency based alignment methods have been designed to: – Better extract the signal contained in the data – Integrate/Confront existing methods – Integrate/Confront heterogeneous types of Information

24. The Progressive Alignment Algorithm

25. T-Coffee and Concistency… SeqA GARFIELD THE LAST FAT CAT SeqB GARFIELD THE FAST CAT SeqC GARFIELD THE VERY FAST CAT SeqD THE FAT CAT SeqA GARFIELD THE LAST FA-T CAT SeqB GARFIELD THE FAST CA-T --- SeqC GARFIELD THE VERY FAST CAT SeqD -------- THE ---- FA-T CAT

26. T-Coffee and Concistency… SeqA GARFIELD THE LAST FAT CAT Prim. Weight =88 SeqB GARFIELD THE FAST CAT --- SeqA GARFIELD THE LAST FA-T CAT Prim. Weight =77 SeqC GARFIELD THE VERY FAST CAT SeqA GARFIELD THE LAST FAT CAT Prim. Weight =100 SeqD -------- THE ---- FAT CAT SeqB GARFIELD THE ---- FAST CAT Prim. Weight =100 SeqC GARFIELD THE VERY FAST CAT SeqC GARFIELD THE VERY FAST CAT Prim. Weight =100 SeqD -------- THE ---- FA-T CAT

27. T-Coffee and Concistency… SeqA GARFIELD THE LAST FAT CAT Prim. Weight =88 SeqB GARFIELD THE FAST CAT --- SeqA GARFIELD THE LAST FA-T CAT Prim. Weight =77 SeqC GARFIELD THE VERY FAST CAT SeqA GARFIELD THE LAST FAT CAT Prim. Weight =100 SeqD -------- THE ---- FAT CAT SeqB GARFIELD THE ---- FAST CAT Prim. Weight =100 SeqC GARFIELD THE VERY FAST CAT SeqC GARFIELD THE VERY FAST CAT Prim. Weight =100 SeqD -------- THE ---- FA-T CAT SeqA GARFIELD THE LAST FAT CAT Weight =88 SeqB GARFIELD THE FAST CAT --- SeqA GARFIELD THE LAST FA-T CAT Weight =77 SeqC GARFIELD THE VERY FAST CAT SeqB GARFIELD THE ---- FAST CAT SeqA GARFIELD THE LAST FA-T CAT Weight =100 SeqD -------- THE ---- FA-T CAT SeqB GARFIELD THE ---- FAST CAT

28. T-Coffee and Concistency… SeqA GARFIELD THE LAST FAT CAT Weight =88 SeqB GARFIELD THE FAST CAT --- SeqA GARFIELD THE LAST FA-T CAT Weight =77 SeqC GARFIELD THE VERY FAST CAT SeqB GARFIELD THE ---- FAST CAT SeqA GARFIELD THE LAST FA-T CAT Weight =100 SeqD -------- THE ---- FA-T CAT SeqB GARFIELD THE ---- FAST CAT

29. T-Coffee and Concistency…

31. Methods Data Scalability

32. A Brief History of Consistency A Long Chain of Small Contributions…

33. Consistency Based Algorithms Gotoh (1990) – Iterative strategy using consistency Martin Vingron (1991) – Dot Matrices Multiplications – Accurate but too stringeant Dialign (1996, Morgenstern) – Concistency – Agglomerative Assembly T-Coffee (2000, Notredame) – Concistency – Progressive algorithm ProbCons (2004, Do) – T-Coffee with a Bayesian Treatment – Biphasic Gap Penalty AMAP (Schwarz, 2007) – ProbCons Consistency – Replace Progressive alignment with simulated Annealing – Hard to distinguish from ProbCons FSA ( Patcher, 2009) – AMAP with automated parameter estimation – Hard to distinguish from ProbCons

34. Choosing the right modeling method M-Coffee

35. Combining Many MSAs into ONE MUSCLE MAFFT ClustalW ??????? T-Coffee

36. Consistency and Accuracy

37. Integrating New Types of Data Template Based Sequence Alignments

38. Experimental Data … TARGET Experimental Data … TARGET Template Aligner Template-Sequence Alignment Primary Library Template Alignment Template based Alignment of the Sequences Templates TARGET

39. Exploring The Template World TemplateGeneratorAlignment Method RNA StructurePredictionRNA Aligner Protein StructureBLAST vs PDB3D Aligner ProfileBLAST vs NRProfile/Profile Alignment Gene StructureENSEMBLGenome Aligner PromoterTransfacMeta-Aligner

40. Exploring The Template World TemplateGeneratorAlignment Method Mode RNA Structure PredictionRNA Aligner R-Coffee Protein Structure BLAST /PDB3D Aligner 3D-Coffee Profile BLAST/NRProfile/Profile PSI-Coffee Gene Structure ENSEMBLGenome Aligner Exoset Promoter TransfacMeta-Aligner Meta-Coffee

41. 3D-Coffee/Expresso Incorporating Structural Information

42. Expresso: Finding the Right Structure Sources Templates Library BLAST SAP Template Alignment Source Template Alignment Remove Templates Templates

43. PSI-Coffee Homology Extension

44. Exploring The Template World

45. What is Homology Extension ? LL L ? -Simple scoring schemes result in alignment ambiguities

46. What is Homology Extension ? LL L L L L L L L L L I V I L L L L L L L Profile 1 Profile 2

47. What is Homology Extension ? LL L L L L L L L L L I V I L L L L L L L Profile 1 Profile 2

48. PSI-Coffee: Homology Extension Sources Templates Library BLAST Template Alignment Source Template Alignment Remove Templates Templates Profile Aligner

49. Benchmarks

50. Method TemplateScoreComment ClustalW-2ProgressiveNO22.74 PRANKGapNO26.18Science2008 MAFFTIterativeNO26.18 MuscleIterativeNO31.37 ProbConsConsistencyNO40.80 ProbConsMonoPhasicNO37.53 T-CoffeeConsistencyNO42.30 M-Coffe4ConsistencyNO43.60 PSI-CoffeeConsistencyProfile53.71 PROMALConsistencyProfile55.08 PROMAL-3DConsistencyPDB57.60 3D-CoffeeConsistencyPDB61.00Expresso Score: fraction of correct columns when compared with a structure based reference (BB11 of BaliBase).

51. Method TemplateScoreComment ClustalW-2ProgressiveNO22.74 PRANKGapNO26.18Science2008 MAFFTIterativeNO26.18 MuscleIterativeNO31.37 ProbConsConsistencyNO40.80 ProbConsMonoPhasicNO37.53 T-CoffeeConsistencyNO42.30 M-Coffe4ConsistencyNO43.60 PSI-CoffeeConsistencyProfile53.71 PROMALConsistencyProfile55.08 PROMAL-3DConsistencyPDB57.60 3D-CoffeeConsistencyPDB61.00Expresso Score: fraction of correct columns when compared with a structure based reference (BB11 of BaliBase). Consistency

52. Method TemplateScoreComment ClustalW-2ProgressiveNO22.74 PRANKGapNO26.18Science2008 MAFFTIterativeNO26.18 MuscleIterativeNO31.37 ProbConsConsistencyNO40.80 ProbConsMonoPhasicNO37.53 T-CoffeeConsistencyNO42.30 M-Coffe4ConsistencyNO43.60 PSI-CoffeeConsistencyProfile53.71 PROMALConsistencyProfile55.08 PROMAL-3DConsistencyPDB57.60 3D-CoffeeConsistencyPDB61.00Expresso Score: fraction of correct columns when compared with a structure based reference (BB11 of BaliBase). Homology Extension

53. Method TemplateScoreComment ClustalW-2ProgressiveNO22.74 PRANKGapNO26.18Science2008 MAFFTIterativeNO26.18 MuscleIterativeNO31.37 ProbConsConsistencyNO40.80 ProbConsMonoPhasicNO37.53 T-CoffeeConsistencyNO42.30 M-Coffe4ConsistencyNO43.60 PSI-CoffeeConsistencyProfile53.71 PROMALConsistencyProfile55.08 PROMAL-3DConsistencyPDB57.60 3D-CoffeeConsistencyPDB61.00Expresso Score: fraction of correct columns when compared with a structure based reference (BB11 of BaliBase). Structural Extension

54. T-Coffee and The World BLAST/ SOAP -Some Templates are obtained with a BLAST -Queries can be sent to the EBI or the NCBI -No Need for a Local BLAST installation Users sequences

55. Incorporating RNA Information Within the T-Coffee Algorithm

56. ncRNAs Can Evolve Rapidly CCAGGCAAGACGGGACGAGAGTTGCCTGG CCTCCGTTCAGAGGTGCATAGAACGGAGG **-------*--**---*-**------** GAACGGACCGAACGGACC CTTGCCTGGCTTGCCTGG G G A A CC A C G G A G A C G CTTGCCTCCCTTGCCTCC GAACGGAGGGAACGGAGG G G A A CC A C G G A G A C G

57. ncRNAs Can Evolve Rapidly CCAGGCAAGACGGGACGAGAGTTGCCTGG CCTCCGTTCAGAGGTGCATAGAACGGAGG **-------*--**---*-**------** CC--AGGCAAGACGGGACGAGAGTTGCCTGG CCTCCGTTCAGAGGTGCATAGAAC--GGAGG ** * *** * * *** ** Sequence Alignment (Maximizing Identity) -Incorrect -Not Predictive

58. The Holy Grail of RNA Comparison: Sankoff’ Algorithm

59. C C R-Coffee Extension G G TC Library G G Score X C C Score Y C C G G Goal: Embedding RNA Structures Within The T-Coffee Libraries The R-extension can be added on the top of any existing method.

60. R-Coffee + Structural Aligners MethodAvg BraliscoreNet Improv. direct +T+R +T+R ----------------------------------------------------------- Stemloc0.620.750.76 104113 Mlocarna0.660.690.71 101133 Murlet0.730.700.72 -132-73 Pmcomp0.730.730.73 142145 T-Lara0.740.740.69 -36 -8 Foldalign0.750.770.77 72 73 ----------------------------------------------------------- Dyalign---0.630.62 ------ Consan---0.790.79 ------ ----------------------------------------------------------- Improvement= # R-Coffee wins - # R-Coffee looses over 170 test sets

61. R-Coffee + Regular Aligners MethodAvg BraliscoreNet Improv. direct +T+R +T+R ----------------------------------------------------------- Poa0.620.650.70 48154 Pcma0.620.640.67 34120 Prrn0.640.610.66-63 45 ClustalW0.650.650.69 -7 83 Mafft_fftnts0.680.680.72 17 68 ProbConsRNA0.690.670.71-49 39 Muscle0.690.690.73-17 42 Mafft_ginsi0.700.680.72-49 39 ----------------------------------------------------------- Improvement= # R-Coffee wins - # R-Coffee looses over 388 test sets

62. Genomic Era Challenges Conclusion Template Based Alignments Meta-Methods M-Coffee Homology Extension (Proteins) R-Coffee Scaled Consistency

63. Open Questions Accurately Aligning non transcribed DNA Coping with One Billion Human Genomes

64. www.tcoffee.org cedric.notredame@crg.es Comparative Bioinformatics University College Dublin – Des Higgins – Orla O’Sullivan – Iain Wallace (UCD, IE) Berlin Free University – Knut Reinert – Tobias Rausch Swiss Intitute of Bioinformatics – Ioannis Xenarios – Sebastien Morreti Comparative Bioinformatics – Merixell Oliva – Giovanni Bussoti – Carsten Kemena – Emanuele Rainieri – Ionas Erb – Jia Ming Chang – Matthias Zytneki

65. www.tcoffee.org cedric.notredame@europe.com www.tcoffee.org

67. Why So Much Interest For Multiple Alignments ? Extrapolation Motifs/Patterns Phylogeny Profiles Structure Prediction SNP Analysis Reactivity Analysis Regulatory Elements

68. Phylogeny Vs Function: Applications Comparative Genomics => New Medium New Medium => New Clinical Test

69. Detecting ncRNAs in silico: a long way to go… RNAse P

70. Obtaining the Structure of a ncRNA is difficult Hard to Align The Sequences Without the Structure Hard to Predict the Structures Without an Alignment

71. R-Coffee: Modifying T-Coffee at the Right Place Incorporation of Secondary Structure information within the Library Two Extra Components for the T-Coffee Scoring Scheme – A new Library – A new Scoring Scheme

74. G-INS-i, H-INS-i and F-INS-i use pairwise alignment information when constructing a multiple alignment. The two options ([HF]-INS-i) incorporate local alignment information and do NOT USE FFT.

75. Molecular Biology Within the System Biology Era Protein A Interacts with Regulates Inhibits Protein B

76. Molecular Biology In the 1000 Genomes Era Protein A Interacts with Regulates Inhibits Protein B Variation Within Species: CNVs of A and SNPs Conservation Across Species

77. System Biology vs Comparative Genomics Systems Biology  Systems can be Understood Comparative Genomics  Systems can Evolve through Selection

78. Phylogeny Vs Function: Applications – Important  Application – Possible  Many New Genomes – Challenging  Too Many New Genomes

79. 3D-Coffee: Combining Sequences and Structures Within Multiple Sequence Alignments

80. Comparing Methods MAFFT

81. Some Benchmark: BB11 BaliBase BB11: 38 highly divergent (less than 25% id) datasets from BaliBase BB11: predicts 78% of the results measured on other datasets Blackshield, Higgins

82. PhD Fellowships www.crg.es

85. What ‘s in a Multiple Sequence Alignment Evolution Inertia Common Ancestry Shows up In the sequences Selection Important Features Are Preserved Functional Constraint Same Function Same Sequence Convergence Phylogenetic Footprint, Evolutionary Trace …

86. Which Tool for Which Sequence ?

87. Is it Possible to Compare all Types of Sequences Non Transcribed World – Genes/Full Genomes Lagan, TBA – Promoter Regions Meta-Aligner Motifs Finders – Nucleosome ??? Multiple Genome Aligners – Not Very Accurate – Very Fast – Deal with rearrangements

88. Is it Possible to Compare all Types of Sequences RNA Comparison – Less Accurate than Proteins – Secondary Structures ncRNA World – Consan – R-Coffee

89. Is it Possible to Compare all Types of Sequences Protein Comparisons – Very Accurate – 3D-Structure Improves it Protein Aligners – ClustalW – T-Coffee – 3D-Coffee

90. Why So Much Interest For Multiple Alignments ? Extrapolation Motifs/Patterns Phylogeny Profiles Structure Prediction SNP Analysis Reactivity Analysis Regulatory Elements

91. What’s in a Multiple Alignment ? The MSA contains what you put inside: – Structural Similarity – Evolutive Similarity – Sequence Similarity You can view your MSA as: – A record of evolution – A summary of a protein family – A collection of experiments made for you by Nature…

92. Producing The Right Alignment Multiple Sequence Alignments Influence Phylogenetic Trees Choice of Method is not Neutral – Different Methods – Different Alignments – Different Trees Using The Right Models insures Producing the right Tree

93. Model Based Alignments vs Naïve Alignments Naïve Alignment – Lexicographic Alignment – Maximizing the number of identities – At best using a substitution matrix Model Based Alignments – Using a model – Protein structure information – RNA Structure information – Combining/Confronting Modeling methods Template based Alignments – Model based Alignments through the use of Templates

94. T-Coffee and Model Based Alignments T-Coffee Algorithm Expresso: Aligning Protein Structures R-Coffee: Aligning RNA structures M-Coffee: Combining methods

95. T-Coffee and Concistency…

96. When Sequences Are not Enough 3D-Coffee and Expresso

97. 3D-Coffee: Combining Sequences and Structures Within Multiple Sequence Alignments

99. Where to Trust Your Alignments Most Methods Agree Most Methods Disagree

100. Conclusion Model Based Alignments Give the best Accuracy Template based alignment is a very efficient way to turn Naïve aligners into model based aligners Sequence Alignments are not necessarily reliable over their entire lengths

101. Manguel M, Samaniego F.J., Abraham Wald’s Work on Aircraft Suvivability, J. American Statistical Association. 79, 259-270, (1984)

102. Building and Using Models 35.67 Angstrom

103. Computing the Correct Alignment is a Complicated Problem

104. Stochastic Optimization

105. Exploration of Complex Optimization Problems With Multiple Constraints – Genomic Alignments – RNA Alignments Generation of Population of Suboptimal Solutions – Quality=f( optimality ) Specification of Concistency Objective Function of T- Coffee

106. Three Types of Algorithms Progressive: ClustalW Iterative: Muscle Concistency Based: T-Coffee and Probcons

107. T-Coffee and Concistency… Each Library Line is a Soft Constraint (a wish) You can’t satisfy them all You must satisfy as many as possible (The easy ones)

108. Concistency Based Algorithms: T-Coffee Gotoh (1990) – Iterative strategy using consistency Martin Vingron (1991) – Dot Matrices Multiplications – Accurate but too stringeant Dialign (1996, Morgenstern) – Concistency – Agglomerative Assembly T-Coffee (2000, Notredame) – Concistency – Progressive algorithm ProbCons (2004, Do) – T-Coffee with a Bayesian Treatment

109. How Good Is My Method ?

110. Structures Vs Sequences

111. T-Coffee Results Validation Using BaliBase

112. Too Many Methods for ONE Alignment M-Coffee

114. Estimating the Accuracy of your MSA

115. What To Do Without Structures

116. 3D-Coffee: Combining Sequences and Structures Within Multiple Sequence Alignments

117. Expresso: Finding the Right Structure Why Not Using Structure Based Alignments

118. Template Based Multiple Sequence Alignments

119. -Structure -Profile -… Sources Templates Library Template Aligner Template Alignment Source Template Alignment Remove Templates Templates -Structure -Profile -…

120. MethodScoreTemplatesPrefabHomstrad -------------------------------------------------------------- ClustalWMatrix----61.80---- KalignMatrix----63.00---- MUSCLEMatrix----68.0045.0 -------------------------------------------------------------- T-CoffeeConsistency ----69.9744.0 ProbConsConsistency----70.54---- MafftConsistency----72.20---- M-CoffeeConsistency----72.91---- MUMMALSConsistency----73.10---- -------------------------------------------------------------- Clustal-db MatrixProfiles-------- PRALINEMatrixProfiles----50.2 PROMALSConsistencyProfiles79.00---- SPEMMatrixProfiles77.00---- -------------------------------------------------------------- EXPRESSOConsistencyStructures----71.9 * T-LaraConsistencyStructures-------- -------------------------------------------------------------- Table 1. Summary of all the methods described in the review. Validation figures were compiled from several sources, and selected for the compatibility. Prefab refers to some validation made on Prefab Version 3. The HOMSTRAD validation was made on datasets having less than 30% identity. The source of each figure is indicated by a reference. *The EXPRESSO figure comes from a slightly more demanding subset of HOMSTRAD (HOM39) made of sequences less than 25% identical.

121. Improving The Evaluation

122. How Do We Perform In The Twilight Zone? Concistency Based Methods Have an Edge Hard to tell Methods Apart Sequence Alignment is NOT solved

123. More Than Structure based Alignments Structural Correctness Is Only the Easy Side of the Coin. In practice MSA are intermediate models used to generate other models: DataModel TypeBenchmark HomologyProfileYes EvolutionTreesNo Structure3D-StructureCASP FunctionAnnotationNo

124. Conclusion Template based Multiple Sequence Alignments Projecting any relevant information onto the sequences Using this Information Need for new evaluation procedures Functional Analysis Phylogenetic Analysis Homology Search (Profiles) Homology Modelling Integrating data  Making sure your bits of data can fight with one another

125. Turning Data into Models Data Columbus, considered that the landmass occupied 225°, leaving only 135° of water (Marinus of Tyre, 70 AD).Marinus of Tyre Columbus believed that 1° represented only 56 miles (Alfraganus, XIth century)Alfraganus He knew there was an island named Japan off the cost of China… Model Circumference of the Earth as 25,255 km at most, Canary Island to Japan : 3,700 km (Reality: 12,000 km.)

126. The More Structures The Merrier Average Improvement over T-Coffee Struc/Seq Ratio

127. The Right Mixt of Methods

128. 3D-Coffee: Combining Sequences and Structures Within Multiple Sequence Alignments

129. Applications

130. Looking-Up The DNA Behind The Sequences: PROTOGENE

131. SAR Analysis Correlate Alignment Variations with Reactivity Application to the Human Kinome Collaboration with Sanofi-Aventis Main Issue: – Training problem  Proper Benchmarking

132. ncRNA Multiple Alignments with R-Coffee Laundering the Genome Dark Matter Cédric Notredame Comparative Bioinformatics Group Bioinformatics and Genomics Program

133. No Plane Today…

134. ncRNAs Comparison And ENCODE said… “nearly the entire genome may be represented in primary transcripts that extensively overlap and include many non-protein-coding regions” Who Are They? – tRNA, rRNA, snoRNAs, – microRNAs, siRNAs – piRNAs – long ncRNAs (Xist, Evf, Air, CTN, PINK…) How Many of them – Open question – 30.000 is a common guess – Harder to detect than proteins.

135. ncRNAs can have different sequences and Similar Structures

136. ncRNAs are Difficult to Align Same Structure  Low Sequence Identity Small Alphabet, Short Sequences  Alignments often Non- Significant

137. Obtaining the Structure of a ncRNA is difficult Hard to Align The Sequences Without the Structure Hard to Predict the Structures Without an Alignment

138. The Holy Grail of RNA Comparison: Sankoff’ Algorithm

139. The Holy Grail of RNA Comparison Sankoff’ Algorithm Simultaneous Folding and Alignment – Time Complexity: O(L 2n ) – Space Complexity: O(L 3n ) In Practice, for Two Sequences: – 50 nucleotides: 1 min.6 M. – 100 nucleotides 16 min.256 M. – 200 nucleotides 4 hours 4 G. – 400 nucleotides3 days3 T. Forget about – Multiple sequence alignments – Database searches

140. The next best Thing: Consan Consan = Sankoff + a few constraints Use of Stochastic Context Free Grammars – Tree-shaped HMMs – Made sparse with constraints The constraints are derived from the most confident positions of the alignment Equivalent of Banded DP

141. Going Multiple…. Structural Aligners

142. Game Rules Using Structural Predictions – Produces better alignments – Is Computationally expensive Use as much structural information as possible while doing as little computation as possible…

143. Adapting T-Coffee To RNA Alignments

144. T-Coffee and Concistency…

148. Consistency: Conflicts and Information X Y X Z Y WZ X Z Y Z W Y W X Z Y Z X W Y Z X W Partly Consistent  Less Reliable Fully Consistent  More Reliable Y is unhappy X is unhappy X Y

149. RNA Sequences Secondary Structures Primary Library R-Coffee Extended Primary Library Progressive Alignment Using The R-Score RNAplfold Consan or Mafft / Muscle / ProbCons R-Coffee Extension R-Score

150. C C R-Coffee Scoring Scheme G G R-Score (CC)=MAX(TC-Score(CC), TC-Score (GG))

151. Validating R-Coffee

152. RNA Alignments are harder to validate than Protein Alignments Protein Alignments  Use of Structure based Reference Alignments RNA Alignments  No Real structure based reference alignments – The structures are mostly predicted from sequences – Circularity

153. BraliBase and the BraliScore Database of Reference Alignments 388 multiple sequence alignments. Evenly distributed between 35 and 95 percent average sequence identity Contain 5 sequences selected from the RNA family database Rfam The reference alignment is based on a SCFG model based on the full Rfam seed dataset (~100 sequences).

154. BraliBase SPS Score RFam MSA Number of Identically Aligned Pairs SPS= Number of Aligned Pairs

155. BraliBase: SCI Score RNApfoldRNApfold (((…)))…((..))  G Seq1 (((…)))…((..))  G Seq2 (((…)))…((..))  G Seq 3 (((…)))…((..))  G Seq4 (((…)))…((..))  G Seq5 (((…)))…((..))  G Seq6 RNAlifold (((…)))…((..)) ALN  G Average  G Seq X Cov  G ALN SCI= Covariance

156. BRaliScore Braliscore= SCI*SPS

157. RM-Coffee + Regular Aligners MethodAvg BraliscoreNet Improv. direct +T+R +T+R ----------------------------------------------------------- Poa0.620.650.70 48154 Pcma0.620.640.67 34120 Prrn0.640.610.66-63 45 ClustalW0.650.650.69 -7 83 Mafft_fftnts0.680.680.72 17 68 ProbConsRNA0.690.670.71-49 39 Muscle0.690.690.73-17 42 Mafft_ginsi0.700.680.72-49 39 ----------------------------------------------------------- RM-Coffee40.71 /0.74 / 84

158. How Best is the Best…. M-Locarna234 ***183 ** Stral169 *** 62 FoldalignM146 61 Murlet130 *-12 Rnasampler129 *-27 T-Lara125 *-30 Poa241 ***217 *** T-Coffee241 ***199 *** Prrn232 ***198 *** Pcma218 ***151 *** Proalign216 ***150 ** Mafft fftns206 ***148 * ClustalW203 ***136 *** Probcons192 ***128 * Mafft ginsi170 ***115 Muscle169 ***111 Method vs. R-Coffee-Consan vs. RM-Coffee4

159. Range of Performances Effect of Compensated Mutations

160. Conclusion/Future Directions T-Coffee/Consan is currently the best MSA protocol for ncRNAs Testing how important is the accuracy of the secondary structure prediction Going deeper into Sankoff’s territory: predicting and aligning simultaneously

161. www.tcoffee.org cedric.notredame@europe.com Credits and Web Servers Andreas Wilm Des Higgins Sebastien Moretti Ioannis Xenarios Cedric Notredame CGR, SIB, UCD

162. Prank Vs Prank

163. Gop=0 Gep=0

164. Prank Vs Prank The reconstruction of evolutionary homology -- including the correct placement of insertion and deletion events -- is only feasible for rather closely-related sequences. PRANK is not meant for the alignment of very diverged protein sequences. If sequences are very different, the correct homology cannot be reconstructed with confidence and http://www.ebi.ac.uk/goldman -srv/prank/

165. Do Benchmarks All Tell the same story? Based on

166. Probcons: Different Primary Library Score=  (MIN(xz,zk))/MAX(xz,zk) Score(xi ~ yj | x, y, z)  ∑k P(xi ~ zk | x, z) P(zk ~ yj | z, y)