Finding a maximum likelihood tree is hard
Journal of the ACM (JACM)
Dynamic multigrain parallelization on the cell broadband engine
Proceedings of the 12th ACM SIGPLAN symposium on Principles and practice of parallel programming
Exploring New Search Algorithms and Hardware for Phylogenetics: RAxML Meets the IBM Cell
Journal of VLSI Signal Processing Systems
Large-scale maximum likelihood-based phylogenetic analysis on the IBM BlueGene/L
Proceedings of the 2007 ACM/IEEE conference on Supercomputing
Large-scale phylogenetic analysis on current HPC architectures
Scientific Programming - Large-Scale Programming Tools and Environments
ZARAMIT: A System for the Evolutionary Study of Human Mitochondrial DNA
IWANN '09 Proceedings of the 10th International Work-Conference on Artificial Neural Networks: Part II: Distributed Computing, Artificial Intelligence, Bioinformatics, Soft Computing, and Ambient Assisted Living
The CIPRES science gateway: a community resource for phylogenetic analyses
Proceedings of the 2011 TeraGrid Conference: Extreme Digital Discovery
Hi-index | 3.84 |
Motivation: Maximum likelihood (ML) is an increasingly popular optimality criterion for selecting evolutionary trees. Yet the computational complexity of ML was open for over 20 years, and only recently resolved by the authors for the Jukes--Cantor model of substitution and its generalizations. It was proved that reconstructing the ML tree is computationally intractable (NP-hard). In this work we explore three directions, which extend that result. Results: (1) We show that ML under the assumption of molecular clock is still computationally intractable (NP-hard). (2) We show that not only is it computationally intractable to find the exact ML tree, even approximating the logarithm of the ML for any multiplicative factor smaller than 1.00175 is computationally intractable. (3) We develop an algorithm for approximating log-likelihood under the condition that the input sequences are sparse. It employs any approximation algorithm for parsimony, and asymptotically achieves the same approximation ratio. We note that ML reconstruction for sparse inputs is still hard under this condition, and furthermore many real datasets satisfy it. Contact: tamirtul@post.tau.ac.il