Inferring phylogenetic trees: Maximum likelihood methods Prof. William Stafford Noble Department of Genome Sciences Department of Computer Science and Engineering University of Washington thabangh@gmail.com One-minute responses • First part of class was fine. • I am struggling with Python. • At first it was difficult to complete the program when I get the first half, but it is getting easier now. • The class lecture is always fine, but the Python problems are getting tougher. However, they are really interesting and quite informative. • We are learning a lot about programming. • The class is more interesting every day. I enjoy the Python, especially because I am able to fill in by myself. • Thank you for helping us with sys.stdout.write. It will be very useful for future work in Python. Outline • Parsimony • Distance methods – Computing distances – Finding the tree • Maximum likelihood Revision Multiple sequence alignment Pairwise distance matrix Phylogenetic tree Revision • Ideally, distances in a phylogenetic tree would represent time. In practice, however, what do the distance estimate represent? – The expected number of changes per position. • What is a “back mutation”? – A pair of mutations that reverse one another (e.g., A C A) Revision • Compute the Juke-Cantor distance between the first yeast and mouse sequences shown below. 3 4 K AB ln1 d AB 4 3 XX X X X dha2_yeast dhac_mouse dha5_yeast dhal_ecoli 93 93 92 92 XX X X X LRYTRHEPVGVCGEIIPWNI FTYTRREPIGVCGQIIPWNI FAYTLKVPFGVVAQIVPWNI LAMIVREPVGVIAAIVPWNI 3 æ 4 5ö K AB = - ln ç1÷ 4 è 3 20 ø 3 æ2ö = - ln ç ÷ 4 è3ø = 0.304 Perform the next merger Spar Smik-Sbay Skud-Scer Scas Sklu Spar 0 31.5 30.5 300 229 Smik-Sbay 31.5 0 34.25 294 223 Skud-Scer 30.5 34.25 0 319.5 248 Scas 300 294 319.5 0 95 Sklu 229 223 248 95 0 Smik Sbay Skud Scer Perform the next merger Spar Smik-Sbay Skud-Scer Scas Sklu Spar 0 31.5 30.5 300 229 Smik-Sbay 31.5 0 34.25 294 223 Skud-Scer 30.5 34.25 0 319.5 248 Scas 300 294 319.5 0 95 Sklu 229 223 248 95 0 Smik Sbay Skud Scer Perform the next merger Skud-ScerSpar Smik-Sbay Skud-ScerSpar Scas Sklu Skud-ScerSpar 0 32.875 0 309.75 238.5 Smik-Sbay 32.875 0 32.875 294 223 Skud-ScerSpar 0 32.875 0 309.75 238.5 Scas 309.75 294 309.75 0 95 Sklu 238.5 223 238.5 95 0 Smik Sbay Skud Scer Extend the corresponding tree Smik-Sbay Skud-ScerSpar Scas Sklu Smik-Sbay 0 32.875 294 223 Skud-ScerSpar 32.875 0 309.75 238.5 Scas 294 309.75 0 95 Sklu 223 2238.5 95 0 Sklu Scas Smik Sbay Spar Skud Scer Maximum parsimony for each possible tree for each column of the alignment compute the parsimony score of the column, given the tree return the tree with the best parsimony score Maximum likelihood for each possible tree for each column of the alignment compute the likelihood of the column, given the tree return the tree with the highest likelihood • Similar to parsimony, but capable of using a model of evolution. • Computationally expensive. • DNAML is the Phylip program for maximum likelihood. FastDNAML is a fast clone. http://evolution.genetics.washington.edu/phylip.html http://iubio.bio.indiana.edu/soft/molbio/evolve/fastdnaml/fastDNAml.html Problem #1 ACGCGTTGGG ACGCGTTGGG ACGCAATGAA ACACAGGGAA + T Pr(column|tree,model) T A G • What is the probability of observing this column, given this tree and an assumed model of evolution? Solution #1 C A A T A A T A G T G A A T A G T A T A G • Solution: Enumerate all possible assignments to the internal nodes. Compute the probability of each tree, and sum. Problem #2 ACGCGTTGGG ACGCGTTGGG ACGCAATGAA ACACAGGGAA + T A T Pr(column|tree,model) A T A G • What is the probability of observing this column, given this assigned tree and an assumed model of evolution? Solution #2 The probability of observing a substitution from A to T on a branch of length m is given by the evolutionary model. πA, πC, πG, πT The probability of the ancestral observation being A is just πA. m A T T A T A G Solution #2 πA, πC, πG, πT L0 A L1 L2 T L5 A L3 L4 L6 T T A G • The desired probability is the product of the probabilities of the branches. • L(tree) = L0 L1 L2 L3 L4 L5 L6 Computing the likelihood C A A T A A T A tree1 G T G A A T A tree2 G T A T A tree3 • The probability of the tree is the sum of the probabilities of the individual trees. • L(tree) = L(tree1) + L(tree2) + L(tree3) + … G Maximum likelihood revisited for each possible tree for each column of the alignment for each assignment of internal nodes for each branch compute the probability of that branch assigned tree probability ← multiply branch probabilities column probability ← sum assigned tree probabilities tree probability ← multiply column probabilities return the tree with the highest probability Maximum likelihood revisited for each possible tree Multiply probabilities of for each column of the alignment independent for each assignment of internal nodes events. for each branch compute the probability of that branch assigned tree probability ← multiply branch probabilities column probability ← sum assigned tree probabilities tree probability ← multiply column probabilities return the tree with the highest probability Add probabilities of mutually exclusive events. Overview • Parsimony • Distance methods – Computing distances – Finding the tree • Fitch-Margoliash • Neighbor-joining • UPGMA • Maximum likelihood Representing trees • ((mouse, rat), (human, chimp)) mouse rat human myTree = [[mouse, rat], [human, chimp]] chimp Problem #1 • Write a program to read a parenthesized tree from a file and count the number of nodes. > cat mytree.txt (yeast, ((fly, spider), (dog, cat))) > python read-tree.py mytree.txt Read 5 species from mytree.txt. Problem #2 • Modify the previous program to print the leaves of the tree, indenting according to the depth. > print-tree.py mytree.txt yeast fly spider dog cat Problem #3 • Given: a three-column file in which the first two columns contain names of species and the third column contains the distance between them. • Print to standard output a formatted matrix in which the species names are listed in the rows and columns, and values are from the input file. – Species should be listed in alphabetical order. – The program should halt and complain if a value is missing. – The matrix is assumed to be symmetric, and each pair appears only once. – Distances of zero along the diagonal are not included in the input. – Columns should be printed in the same width as the corresponding species name. ./print-distance-matrix.py distances.txt Read 30 values and 6 species from distances.txt. Maximum species name width = 9. ape cat dog gerbil mouse zebrafish ape 0 0.19 0.15 0.44 0.17 0.69 cat 0.19 0 0.1 0.48 0.24 0.77 dog 0.15 0.1 0 0.43 0.25 0.78 gerbil 0.44 0.48 0.43 0 0.42 0.78 mouse 0.17 0.24 0.25 0.42 0 0.85 zebrafish 0.69 0.77 0.78 0.78 0.85 0