UVM Theses and Dissertations
Format:
Print
Author:
Barrantes-Reynolds, Ramiro
Dept./Program:
Cell and Molecular Biology Program
Year:
2011
Degree:
Ph. D.
Abstract:
Structural features of enzymes appear and disappear throughout evolution; for example, a zinc finger or a salt bridge may be present in some, but not all, members of a protein family. However, current methods do not explicitly identify these features. We propose that such events 1) are important sources of functional divergence in enzymes, 2) can be identified using sequence alignments and one structure, and 3) evolve more slowly than individual amino acids, thus making them useful for phylogenetic inference of ancient protein subfamilies.
We developed a Bayesian statistical model that both simultaneously identifies structural characters whose states are motifs that change in selection pressure across the sub members of a protein family; and infers the phylogeny of ancient subfamilies using such characters. The model is based on partitioning the alignment into blocks of neighboring positions based on the 3D structure, where each block evolves independently by allowing all of its positions to switch in unison between two states of mutation rates (slow/fast). We used as input the mutation rate of each site across the different subfamilies, and the 3D structure and Markov Chain Monte Carlo (MCMC) for the inference.
We found several structural characters that changed in evolution across nine subfamilies of the Fpg/Nei base excision repair protein family. The model converged for all the desired properties, and shows that multiple sites cooperate in function and span a small proportion of the protein. It also provides a phylogeny of the ancient subfamilies based on mutation rate, and provides the rate-shift sites and groups of sites that changed on each edge.
Our method considers the context of amino acids and finds meaningful structural features that change in rate, as well as a phylogenetic inference, and suggests interesting questions about the way enzymes work, therefore providing a step forward in taking into account structural context in the understanding of enzymes and their evolution.
We developed a Bayesian statistical model that both simultaneously identifies structural characters whose states are motifs that change in selection pressure across the sub members of a protein family; and infers the phylogeny of ancient subfamilies using such characters. The model is based on partitioning the alignment into blocks of neighboring positions based on the 3D structure, where each block evolves independently by allowing all of its positions to switch in unison between two states of mutation rates (slow/fast). We used as input the mutation rate of each site across the different subfamilies, and the 3D structure and Markov Chain Monte Carlo (MCMC) for the inference.
We found several structural characters that changed in evolution across nine subfamilies of the Fpg/Nei base excision repair protein family. The model converged for all the desired properties, and shows that multiple sites cooperate in function and span a small proportion of the protein. It also provides a phylogeny of the ancient subfamilies based on mutation rate, and provides the rate-shift sites and groups of sites that changed on each edge.
Our method considers the context of amino acids and finds meaningful structural features that change in rate, as well as a phylogenetic inference, and suggests interesting questions about the way enzymes work, therefore providing a step forward in taking into account structural context in the understanding of enzymes and their evolution.