700 likes | 1.4k Views
Coffee Shop. F91921025 黃仁暐 F92921029 戴志華 F92921041 施逸優 R93921142 吳於芳 R94921035 林與絜. Menu. Coffee Shop Opening Why coffee shop? Three Flavors COFFEE T-Coffee 3DCoffee Remarks Recipes. Multiple Sequence Alignment.
E N D
Coffee Shop F91921025 黃仁暐 F92921029 戴志華 F92921041 施逸優 R93921142 吳於芳 R94921035 林與絜
Menu • Coffee Shop Opening • Why coffee shop? • Three Flavors • COFFEE • T-Coffee • 3DCoffee • Remarks • Recipes
Multiple Sequence Alignment • Multiple sequence alignment is one of the most important tool for analyzing biological sequence. • structure prediction • phylogenetic analysis • function prediction • polymerase chain reaction (PCR) primer design.
Multiple Sequence Alignment • However, the accuracy is not good enough. • difficult to evaluate the quality of a multiple alignment • algorithmically very hard to produce the optimal alignment • In order to increase the accuracy of multiple sequence alignment, we opened a coffee shop to share three kinds of coffee.
Before (drinking) COFFEE • For comparative genomics, and why? • Understanding the process of evolution at gross level and local level • Translate DNA sequence data into proteins of known function • Meaning of conservative regions • E. coli, C. elegans, Drosophila, Human… • What’s their relationship?
大腸桿菌 線蟲 集胞藻屬(藍綠藻類) 果蠅 人類 酵母菌 阿拉伯芥 Classification for genes of different function Adapted from “Principles of genome analysis and genomics” Fig. 7.5 (p.129), by S. B. Primrose and R. M. Twyman, 3rd edition
Comparative genomics vs. multiple sequence alignment • Alignment → conservative region • Conservative region → gene location • Evolution evidence http://www.public.iastate.edu/~semrich/compgen/
A: human chromosome IB: human chromosome IIC: human chromosome III Chromosome III region 125-128 Mb was magnified 120X The alignment between the chromosomes http://gchelpdesk.ualberta.ca/news/02jun05/cbhd_news_02jun05.php
Our Flavors • COFFEE: A New Objective Function For Multiple Sequence Alignmnent. • C. Notredame, L. Holme and D.G. Higgins,Bioinformatics,Vol 14 (5) 407-422,1998 • T-Coffee: A novel method for multiple sequence alignments. • C.Notredame, D. Higgins, J. Heringa,Journal of Molecular Biology,Vol 302, pp205-217,2000 • 3DCoffee: Combining Protein Sequences and Structures within Multiple Sequence Alignments. • O. O'Sullivan, K Suhre, C. Abergel, D.G. Higgins, C. Notredame. Journal of Molecular Biology,Vol 340, pp385-395,2004
COFFEE • An objective function for multiple sequence alignments • Cédirc Notredame, Liisa Holm and Desmond G. Higgins • SAGA with COFFEE score
Introduction • COFFEE - Consistency based Objective Function For alignmEnt Evaluation • An objective function, COFFEE score, is proposed to measure the quality of multiple sequence alignments • Optimize the COFFEE score of a multiple sequence alignment with the genetic algorithm package SAGA (Sequence Alignment Genetic Algorithm)
Overview of their method • Given • a set of sequences to be aligned • a library containing all pairwise alignments between them, • the COFFEE score reflects the level of consistency between a multiple sequence alignment and the library.
- N 1 N å å × W SCORE ( A ) i , j i , j = = + i 1 j i 1 = COFFEE score - N 1 N å å × W LEN ( A ) i , j i , j = = + i 1 j i 1 with : = SCORE ( A ) number of aligned pairs of residues , i j that are shared between A and the library , i j COFFEE score
Using COFFEE in SAGA • Iteratively, a multiple sequence alignment with higher COFFEE score is generated by SAGA until the COFFEE score cannot be improved • SAGA follows the general principle of genetic algorithm. • The notion of survival of the fittest • SAGA iteratively does: • Evaluate the score of the alignments • The fitter an alignment, the more likely it is to survive and produce an offspring • Alignments survived may be kept unchanged, randomly modified (mutation), or combined with another alignment (cross-over)
SAGA Results COFFEE function COFFEE score & alignment accuracy Optimization of COFFEE function 等下會看到一堆表格很枯燥,所以請忍耐… Effect of optimization Comparison: COFFEE and others Others: PRRP, Clustal W, PILEUP, SAGA MSA, SAM
Optimization • COFFEE function was optimized by SAGA Using SAGA alignments Using ClustalW alignments
Comparison • Multiple alignments of SAGA COFFEE and 5 other methods • PRRP, ClustalW, PILEUP, SAGA MSA, SAM • Performance of SAGA and ClustalW • Comparison of other 5 methods • 即使SAGA-COFFEE不是最好的結果 →跟最好的也相去不遠 • Identity level lower → better SAGA-COFFEE results
Better than PRRP Correctly aligned ratio Worse than PRRP • Ratio of (E+H) residue correctly aligned • Better of worse alignment? SAGA-COFFEE & others • NO such thing as an ideal method
E+H accuracy (%) E+H accuracy (%) r=0.65 Average identity (%) Coffee sequence score COFFEE score and alignment accuracy >85%的sequence都可預測 (error ~ ±10%) 由coffee score去預測alignment的準確度 Average identity 並沒有辦法預測alignment的準確度
Correlation between score and accuracy • Higher score → higher accuracy • SAGA produces more high-score sequence than ClustalW
T-Coffee • A novel method for multiple sequence alignments • C.Notredame, D. Higgins, J. Heringa • ClustalW with extended library
ClustalW ClustalW is the core alignment stradegy of T-Coffee, it follows the procedure below: • Pairwise Alignment: calculate distance matrix • Guide Tree • Unrooted Neighbor-Joining Tree • Rooted Neighbor-Joining Tree: guide tree with sequence weights • Progressive Alignment: align following the guide tree
Guide tree • Use Neighbor-Joining Method to build guide tree from distance matrix. • First construct an unrooted Neighbor-Joining tree, then convert it to a rooted Neighbor-Joining tree, the guide tree.
Progressive Alignment: align following the guide tree Seq5 Seq3 Seq4 Seq1 Seq2 Alignment 2 Alignment 1 Final alignment Alignment 3
Progressive-alignment strategy • Pros • Faster and saving spaces. (compared with computing all possible multiple alignments) • Cons • May not find optimum solution. • Errors made in the rest alignments cannot be rectified later as the rest of the sequences are added in. T-Coffee is an attempt to minimize that effect! “Once a gap, always a gap!”
T-Coffee Algorithm • Generating a primary library of alignments • Derivetion of the primary library weights • Combination of the libraries • Extending the library • Progressive alignment strategy
Lalign Primary Library (Local) ClustalW Primary Library (Global) Weighting Primary Library
Lalign Primary Library (Local) ClustalW Primary Library (Global) Weighting Primary Library Extension Extended Library
A Extended Library Weight(A-C-B) = min( Weigh(A-C), Weight(B-C) ) = min( 77, 100 ) = 77 Weight(A-D-B) = min( Weight(A-D), Weight(B-D) ) = min( 100, 100 ) = 100
SeqA: GARFIELD THE LAST FAT CAT SeqB: GARFIELD THE FAST CAT SeqA: GARFIELD THE LAST FAT CAT A SeqB: GARFIELD THE FAST CAT Extended Library
SeqA: GARFIELD THE LAST FAT CAT SeqB: GARFIELD THE FAST CAT SeqA: GARFIELD THE LAST FAT CAT A SeqB: GARFIELD THE FAST CAT Extended Library
Lalign Primary Library (Local) ClustalW Primary Library (Global) Weighting Primary Library Extension Extended Library Progressive Alignment Multiple Alignment Information
Complexity Analysis • complexity of the whole procedure: O(N2L2) + O(N3L) + O(N3) + O(NL2) • O(N2L2): computation of the pair-wise library • O(N3L): computation of the extended pair-wise library • O(N3): computation of the NJ tree • O(NL2): computation of the progressive alignment • N sequences that can be aligned in a multiple alignment of length L
Experiment • Implementation environment • Result 1: Effect of combining local and global alignments without extension; effect of the library extension • Result 2: compared with other multiple sequence alignment methods
Implementation environment • Programming language: ANSI C • Hardware: LINUX platform with Pentium II processors (330 MHz). • Test case: BaliBase database of multiple sequence alignment
Result 1 Table 1: The effect of combining local and global alignments Name global/local/extend Cat1(81) Cat2(23) Cat3(4) Cat4(12) Cat5(11) Total(141) Significance C ClustalW pw /.../... 70.6 26.7 43.0 56.0 60.0 58.9 7.8 CE ClustalW pw/…/ex 77.1 33.6 47.6 64.8 75.9 66.3 17.7 L .../Lalign pw/... 65.4 12.1 22.8 53.9 66.0 52.0 7.8 LE .../Lalign pw/ex 72.6 25.6 47.2 77.5 85.5 64.2 16.3 CL ClustalW pw/Lalign pw/.. 76.2 32.0 48.3 76.2 74.6 66.5 12.1g CLE ClustalW pw/Lalign pw /ex 80.6 37.1 52.9 83.2 88.6 72.0
Result 2 Table 2: T-coffee compared with other multiple sequence alignment methods Method Cat1(81) Cat2(23) Cat3(4) Cat4(12) Cat5(11) Total1(141) Total2(141) Significance Dialign 71.0 25.2 35.1 74.7 80.4 61.5 57.3 11.3 ClustalW 78.5 32.2 42.5 65.7 74.3 66.4 58.6 26.2 Prrp 78.6 32.5 50.2 51.1 82.7 66.4 59.0 36.9 T-Coffee 80.6 37.1 52.9 83.2 88.6 72.0 68.6
3DCoffee • Combining protein sequences and structures within multiple sequence alignments • O. O'Sullivan, K Suhre, C. Abergel, D.G. Higgins, C. Notredame • T-Coffee with structure information
3DCoffee • Structural information can help to improve the quality of multiple sequence alignments • 3DCoffee • Combines protein sequences and structures • Is based on T-Coffee version 2.00 • Uses a mixture of pairwise sequence alignments and pairwise structure comparison methods.