Functional classification of genes using semantic distance and fuzzy clustering approach: evaluation with reference...

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Gene functional classification groups genes by molecular function or biological process.
• Validating unsupervised gene classification is challenging due to diverse algorithms and distance measures.

Purpose of the Study:

• To evaluate gene functional classification accuracy using KEGG pathways and Pfam clans as ground truth.
• To compare the IntelliGO semantic distance with clustering methods against the DAVID functional classification tool.
• To propose a novel method for discovering missing gene information.

Main Methods:

• Utilized KEGG and Pfam as reference sets for Gene Ontology (GO) biological process and molecular function annotations.
• Employed the F-score method to estimate overlaps between gene clusters and reference sets.
• Tested IntelliGO semantic distance with hierarchical and fuzzy C-means clustering, comparing results with DAVID.

Main Results:

• IntelliGO semantic distance combined with clustering methods showed comparable or improved performance against DAVID.
• The F-score method effectively estimated the overlap between gene clusters and established biological pathway databases.
• Analysis of cluster-reference set matching identified potential for discovering previously unannotated gene functions.

Conclusions:

• The IntelliGO semantic distance offers a robust approach for gene functional classification validation.
• Comparing cluster overlaps with reference sets is a viable strategy for assessing classification accuracy.
• A set-difference approach can effectively identify gaps in current gene annotation knowledge.