Inferring Protein Function from Structure

A major goal of molecular biology is to understand functions of all genes in nature. Accordingly, it is of great importance to improve large-scale functional genomics and proteomics experiments. However, due to costly and time-consuming nature of experiments, bioinformatics approach to infer the function appears to be very attractive. Besides this, there are many proteins of known structure which are not yet functionally characterized. This makes the investigation of sequence-function and structure-function relationships even more necessary. The number of methods for in silico annotation of function has increased enormously over the past few decades, from methods that rely on high sequence similarity between a protein of unknown function and a family of well-characterized proteins to methods that rely on "profiles" to infer the function. Although computational approach of inferring protein function is an important challenge, there are many obstacles to overcome. First, a function is not well defined and can be defined at several levels of detail. Accordingly, it is very difficult to create controlled vocabularies. Second, the precise values for thresholds of significant sequence similarity are actually specific to particular aspects of function and have to be re-established for any given task. The most common approach to study the function is through evolutionary relationship, or homology, with proteins of known function and it is based on the assumptions that "homologous proteins that have similar sequences and structures, have similar functions" which is the so called Sequence-Structure-Function Paradigm. In this research project, the limitations of this approach are studied