Protein Function Prediction and Human Readable Descriptions

Often, the first contact with a protein sequence is its description in a BLAST result. Thus, it is important for genome annotation projects to provide informative, concise and correct descriptions that will give useful hints on a protein's function. Here we present a tool that assigns human readable descriptions to unknown protein sequences. AHRD scores BLAST hits taken from searches against different databases on the basis of the trust put into these databases and the local alignment quality. The BLAST hit descriptions are tokenized into informative words and a lexical analysis scores these tokens according to their frequency and the quality of the BLAST hits they occur in. Shared tokens with Gene Ontology annotations increase the description scoring in order to use standard nomenclature where possible. Finally the best scoring description is assigned.

We have evaluated the performance of AHRD on different test sets and optimized parameters. We also report on the use of domain architecture similarity scores computed from Interproscan results and its effect on selecting correct function descriptions.

In addition, we have implemented a workflow to assign Gene Ontology terms to proteins for precise functional classification. This also combines sequence similarity with domain architecture similarity scoring. Phylogenomic trees are constructed and reconciled with domain architectures. The resulting trees are used as the topology for a Bayesian network to transfer experimentally verified GO annotations.