Title:Fast reconstruction of phylogenetic trees using locally-sensitive hashing

Abstract:As sequence databases increase in size, there is increasing interest in phylogeny algorithms that are both fast and accurate. In response, researchers have developed sub-quadratic time heuristics for building trees from very large alignments \cite{fasttree,WABI}. These heuristics lack theoretical performance guarantees, so it is unclear whether their use is universally appropriate. We present a fast algorithm that correctly reconstructs a large class of phylogenies, with high probability, assuming a Markov model of evolution. The running time of the algorithm is $O(n^{1+\gamma(g)}\log^2n)$, where $\gamma$ is an increasing function of the upper bound on branch lengths in the phylogeny and $\gamma(g)<1$ for all $g$. For phylogenies with short branches, the running time of our algorithm is close to linear. For example, if all branch lengths correspond to a mutation probability of less than 0.02, the running time of our algorithm is roughly $O(n^{1.2}\log^2n)$. Our results demonstrate that many large phylogenies can be reconstructed fast, without compromising the reconstruction accuracy.