Gene annotations prediction using singular value decomposition and independent component analysis

Knowledge of genes features is paramount to understand the biological processes, promising new field of medical research. Analyze genes functions achieved through vocabularies constructed as ontologies, i.e., the association between genes and biomolecular annotations. In recent years, several databases of such annotations have become available, but most are incomplete and include errors, the only subset of them is reliable and reviewed by bioscientists. In this research, a new computational technique applies truncated singular value decomposition (trunSVD) and Independent Component Analysis (IndCA) to the existing biomolecular databases, to quicken revision of the annotations and suggests new reliable ones. Discover probable inconsistencies in existing annotations, missing annotations, and recommends them to the bioscientists as a suggested annotation. Experimental results show that the suggested technique outperforms existing techniques in terms of accuracy. Reconstruction of gene annotations predicted terms hierarchical relationships gives a greater number of discovered annotations, making the suggested technique a valuable tool to select the most reliable predicted annotations.