ISOGO: Functional annotation of protein-coding splice variants.

Area of Science:

• Genomics and Bioinformatics
• Molecular Biology
• Computational Biology

Background:

• RNA-sequencing (RNA-seq) has shifted genetic analysis to the transcriptome.
• Alternative splicing creates diverse gene isoforms, but their specific functions are often unknown.
• Current Gene Ontology (GO) annotations lack isoform-specificity, hindering functional understanding.

Purpose of the Study:

• To develop a novel algorithm, ISOGO (ISOform + GO function imputation), for predicting the functions of coding gene isoforms.
• To improve the accuracy and scope of isoform-specific functional annotations.
• To provide a publicly accessible tool for exploring these predictions.

Main Methods:

• ISOGO integrates protein domain information with correlated expression patterns across 11,373 cancer patients.
• The algorithm predicts Gene Ontology functions for individual coding isoforms.
• Predictions were validated against known isoform-specific functions and benchmarked with the CAFA3 challenge.

Main Results:

• ISOGO achieved an area under the precision-recall curve (AUPRC) five times larger than previous methods.
• The median area under the receiver operating characteristic curve (AUROC) for gene functions was 0.82.
• ISOGO predictions aligned with literature for genes like BRCA1, MADD, VAMP7, and ITSN1, and confirmed the main isoform's role in 99.4% of cases.

Conclusions:

• ISOGO is a highly effective algorithm for imputing isoform-specific Gene Ontology functions.
• The approach significantly advances the ability to functionally annotate diverse gene products.
• A web application and code are available, facilitating broader scientific use and discovery.