This work aims to identify protein functional determinants and to compare them across the proteome to predict protein function. The approach is predicated on a phylogenomic algorithm, the Evolutionary Trace (ET), that identifies key functional residues in proteins;and on ET Annotation (ETA) algorithms, which extract from ET analysis 3D templates, describing the composition and conformation of key residues involved in binding or in catalysis, and then search in other structures for geometric matches to these 3D templates that suggest a common function. Preliminary data have extensively validated ET, both computationally and through experiments, and ETA has become a useful tool to annotate function on structural genomics proteins. Both methods, however, can still gain in sensitivity, specificity and scalability. To do so we propose in Aim 1, first, to improve the ET identification of key functional residues, by optimizing the selection of the input sequences and by a new measure of residue functional importance, and, second, to refine the selection of 3D templates.
In Aim 2, we propose a new network-based annotation diffusion method to compare all 3D template matches at once and to add in functional information from other sources, such as from proteins without known structure.
Aim 3 is experimental and it will test our predictions through mutations and assays on proteins of direct medical interest including one that controls drug resistance in bacteria and another that is a marker of drug resistance in malaria. In the long term, these results should help to focus protein engineering and drug design to the most functionally and therapeutically relevant parts of a protein, and, most broadly, link the massive and exponentially growing amounts of raw sequence and structure data to biological function and its molecular basis.
Modern biology excels at producing volumes of basic information on the composition of our genes and on the structure of the proteins that they encode. However, much of this potentially useful information lies fallow and does not contribute to our understanding of the basic biology of disease or to the development of new drugs and treatments. The reason is that it remains difficult to know what these new genes actually do, and how they do it. This work develops computational methods to answer both questions. In so doing it should help identify the function of novel protein and help connect them to pathological processes. For example, to test some of our tools and predictions, we will experimentally study two proteins of medical interest, one that orchestrates drug resistance in bacteria, and another that marks drug resistance in malaria.
Showing the most recent 10 out of 68 publications
