Inferring Phylogeny using Permutation Patterns on Genomic Data 1 Md Enamul Karim 2 Laxmi Parida 1 Arun Lakhotia 1 University of Louisiana at Lafayette 2 IBM T. J. Watson Research Center
Phylogeny Reconstruction of the evolutionary relationship of a collection of organisms, usually in the form of a tree.
Phylogenetic data Behavioral, morphological, metabolic, etc. Molecular data: sequence data, gene-order data etc. gene-order data
Why gene order data? Low error rate. Rare evolutionary events unlikely to cause “silent" changes; can help inferring millions of years.
Genomes rearrangements Inverted Transposition –7 –6 –5 –4 10 Inversion –8 –7 –6 – Transposition
Breakpoint distance Breakpoints are number of adjacencies present in one genome, but not in the other –3 – For some datasets, a close-to-linear relationship between the breakpoints and evolutionary events may exist. Can be used for building phylogeny (Blanchette et al.).
Limitations of breakpoint The number of breakpoints created by a certain number of inversions may vary. Also, transpositions generally create more breakpoints than inversions. Computing the breakpoint phylogeny is NP-hard.
MPBE (Maximum Parsimony on Binary Encoding) A heuristic for the breakpoint phylogeny (Cosner et al. ). All ordered pairs of signed genes appearing consecutively are coded as binary features. Exponential time complexity, however, much faster than BPAnalysis.
Limitations May fail to find feasible solutions to the breakpoint phylogeny problem.
Observation: The closer is the evolution history, the more permutations (of different granularity) are in common –8 –7 –6 –5 – –3 –2 –7 –6 –5 –4 9 10
Maximal pi-pattern (Eres et al.) Matches permutations at different granularity. Polynomial time complexity.
pi-pattern Example : For S = and k=2 All pi-patterns are: ac, bc, abc, abcc acbcabacbcab abc Pattern with minimum k permutations
Cover P1 covers P2=> Every P1 has a P2 Every P2 is within a P1 Example In S = acbcab abc covers ac
Maximal pi-pattern pi-pattern which is not covered Example In S = acbcab pi-patterns: ac, bc, abc, abcc Maximal pi-patterns: abc, abcc not covered by abcc
Results
Phylogeny for simulated evolution on synthetic data
12 genera of Campanulaceae and the outgroup tobacco
Tree1: MPBE tree
Tree2: Neighbor joining tree (using few different distances)
Tree3: Neighbor joining tree using permutation patterns 167 Maximal pi-patterns(from pi-patterns) used as binary feature XOR Distance measure Distance/Similarity matrix is created to find neighbor joining tree
Tree3 vs Tree2
Conclusion Permutation patterns may preserve more evolutionary information. Evolutionary events could be counted within permuted segments to develop a hybrid scheme. Current approaches remain unable to handle unequal gene content, which could be solved using maximal pi-patterns.