| The completed sequencing of the human genome presents a problem: the tens of thousands of predicted genes, encoding possibly four times as many proteins, greatly outnumber the few thousand proteins with experimentally well-characterized biochemical functions and cellular roles. For the overwhelming majority of these novel proteins functional roles are inferred from similar, already characterized sequences. Traditional functional assignment, based on simple sequence similarity metrics, cannot identify sequence homologs that diverge too greatly. This has led to an inheritance problem as many multiple-domain proteins have evolved by combinatoric fusion of domain modules. Thus assignments based on similarity to a single domain can be greatly misleading.; This work presents a system for assigning functions to discrete domains within unknown, predicted proteins using Bayesian prior-based profiles. I have applied it to the phylogeny and domain dissection of the yeast mitochondrial proteome, and to the analysis of proteins involved in the pathology of human mitochondrial diseases. The system consists of novel software for sequence alignment, profile construction, organelle database management, parallel phylogenetic methodologies and multiple interface presentation.; More than 430 genes from Saccharomyces cerevisiae encode proteins residing in mitochondria or contributing to mitochondrial biosynthesis and/or function. Profiles constructed for these mitoproteins identified thousands of homologs, and profile-induced multiple alignments yielded a wealth of phylogenetic information. Using only protein domains unequivocally alignable across multiple taxa, I have shown that two-thirds of the phylogenetic trees indicate bacterial ancestry (consistent with derivation from a bacterial endosymbiont), but other cases strongly suggest archaeal or eukaryotic origins.; Profile analysis of mitochondrial disease related genes has been performed for 19 mutations involved in Leber's hereditary optic neuropathy (LHON), Leigh syndrome, mitochondrial neurogastrointestinal encephalomyopathy (MNGIE), Mohr-Tranebjaerg syndrome (MTS), iron-storage disorders related to Friedreich's ataxia, and hereditary spastic paraplegia (HSP). Of the 19 mutations analyzed, 14 involved changes in regions covered by profile analysis. Five out of seven structural correlations provided reasonable explanations for the malfunctions. This methodology can be extended to assist in identifying new disease-related genes, and to generate mechanistic hypotheses concerning the underlying molecular defects. |
