Gene function prediction based on Gene Ontology Hierarchy Preserving Hashing.

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Gene Ontology (GO) provides structured vocabularies for gene product functions.
• Accurate GO annotation prediction is challenging due to the vast number of GO terms.
• Existing methods struggle with the scale and hierarchical nature of GO.

Purpose of the Study:

• To develop a semantic method for accurate gene function prediction.
• To address the challenge of predicting GO annotations from massive GO term sets.
• To improve interspecies gene function prediction and BLAST-based approaches.

Main Methods:

• Introduced Gene Ontology Hierarchy Preserving Hashing (HPHash).
• Measured taxonomic similarity between GO terms.
• Employed hierarchy preserving hashing for encoding GO terms into binary codes.
• Projected gene-term association matrix into a low-dimensional space for semantic similarity prediction.

Main Results:

• HPHash demonstrated superior performance in interspecies gene function prediction for Homo sapiens, Mus musculus, and Rattus norvegicus.
• The method proved robust concerning the number of hash functions used.
• Integrating HPHash as a plugin significantly improved BLAST-based gene function prediction accuracy.

Conclusions:

• HPHash provides an effective semantic approach for gene function prediction.
• The method successfully handles the hierarchical structure and scale of GO terms.
• HPHash offers a significant advancement for both standalone and integrated gene function prediction tools.