Isoform function prediction based on bi-random walks on a heterogeneous network.

Area of Science:

• Computational Biology
• Bioinformatics
• Genomics

Background:

• Alternative splicing generates diverse protein isoforms, each executing specific biological functions.
• Understanding isoform-specific functions is vital for developmental biology and cancer research.
• Current functional genomic databases lack isoform-level annotations, hindering research.

Purpose of the Study:

• To develop a computational method, IsoFun, for predicting gene isoform functions.
• To address the challenge of distinguishing functional annotations between gene isoforms.
• To provide accurate functional predictions essential for understanding disease mechanisms.

Main Methods:

• Constructed an isoform functional association network using RNA-seq expression profiles.
• Built a heterogeneous network integrating Gene Ontology (GO) annotations, gene interactions, and gene-isoform relationships.
• Applied a tailored bi-random walk algorithm on the heterogeneous network for GO term-isoform association prediction.

Main Results:

• IsoFun significantly improved prediction accuracy, outperforming state-of-the-art methods.
• Achieved a 17% increase in Area Under the Receiver-Operating Curve (AUROC) and a 44% increase in Area Under the Precision-Recall Curve (AUPRC) at the gene level.
• Successfully differentiated the functions of specific isoforms for genes ADAM15 and BCL2L1 in validation studies.

Conclusions:

• IsoFun provides a robust framework for predicting gene isoform functions.
• The method effectively leverages network-based approaches for functional genomic analysis.
• IsoFun's ability to distinguish isoform functions offers valuable insights into gene regulation and disease.