Predicting Gene Ontology Functions from ProDom and CDD Protein Domains

Abstract

A heuristic algorithm for associating Gene Ontology (GO) defined molecular functions to protein domains as listed in the ProDom and CDD databases is described. The algorithm generates rules for function-domain associations based on the intersection of functions assigned to gene products by the GO consortium that contain ProDom and/or CDD domains at varying levels of sequence similarity. The hierarchical nature of GO molecular functions is incorporated into rule generation. Manual review of a subset of the rules generated indicates an accuracy rate of 87% for ProDom rules and 84% for CDD rules. The utility of these associations is that novel sequences can be assigned a putative function if sufficient similarity exists to a ProDom or CDD domain for which one or more GO functions has been associated. Although functional assignments are increasingly being made for gene products from model organisms, it is likely that the needs of investigators will continue to outpace the efforts of curators, particularly for nonmodel organisms. A comparison with other methods in terms of coverage and agreement was performed, indicating the utility of the approach. The domain-function associations and function assignments are available from our website http://www.cbil.upenn.edu/GO.An important early step in the postgenomic era is the characterization of the biochemical functions of gene products. Accurate computational predictions are a useful resource for both the community at large and the curators that eventually assign function to gene products. The Gene Ontology (GO) (Gene Ontology Consortium 2001) is an ontology, that is, a database of agreed-to terms for molecular functions, biological processes and cellular components. GO also includes relationships between terms such as specialization or part-whole relations. GO was developed to facilitate effective use of this information. We present an automatic method for leveraging curated GO function annotation of proteins to associate GO terms with protein domains that can then be applied to proteins that contain any of the domains.We make the assumption that the functions that a protein is capable of performing are determined by the protein domains that it contains. We use the simplest possible model of this kind; each domain contributes a function independently of any other domain in the protein. The basis of the algorithm, illustrated in Figure 1, is to identify, using an intersection procedure, the GO functions common to a set of proteins that each contain a domain. We determine whether or not a protein has a domain using sequence similarity. As part of the process of associating GO functions with a domain, we determine a p-value threshold for each domain that indicates the level of similarity needed to confer function. As demonstrated by Hegyi and Gerstein (2001), this is an important consideration. We contrast this approach with the comparison of entire proteins where similarities may be found between regions that are not responsible for the transferred annotation (Myers et al. ...

Functional Genomics of the Endocrine Pancreas

“…The assignments of GO functions to the proteins represented by the cDNA clones in the core pancreas clone set (Fig. 1) were made computationally using an algorithm associating the translated protein domains with GO functions (25). Briefly, to assign GO function(s) to the translated sequences, it is assumed that the sequence of a previously characterized functional domain always functions as characterized.…”

Gene expression profiling in human insulinoma tissue: genes involved in the insulin secretion pathway and cloning of novel full-length cDNAs.

“…The usefulness of the GO ontologies was confirmed repeatedly in the annotations of gene expression data, especially after these had been clustered by similarities in patterns of gene expression (Spellman et al 1998). The assignment of GO functions to the proteins represented by the cDNA clusters of insulinoma was performed computationally using an algorithm associating the translated protein domains with GO functions (Schug et al 2002). The distributions of the top-level GO function assignments for the cDNA clusters corresponding to known genes of insulinoma are shown in Table 3.…”

OrthoMCL: Identification of Ortholog Groups for Eukaryotic Genomes

“…upenn.edu/gene-family), as indicated in Figure 4. The report for each selected ortholog group provides an interactive graphical display of the similarity relationships among related sequences, a table of information related to each individual member of the group (including, e.g., functional assignments based on Gene Ontologies; Ashburner et al 2000; The Gene Ontology Consortium 2001; Schug et al 2002), and a multiple sequence alignment of the group members generated using CLUSTALW (Thompson et al 1994). The graphical representation can be customized to selectively show two-way best matches, one-way best matches, and "recent" paralog relationships among group members or all of the related sequences (see Fig.…”