Large-scale automated function prediction of protein sequences and an experimental case study validation on PTEN...

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Machine learning and increased computing power enable advanced protein sequence and transcript variation analysis.
• Automated functional annotation of proteins is crucial for understanding biological systems.

Purpose of the Study:

• To introduce UniGOPred, an automated system for predicting Gene Ontology (GO) annotations for protein sequences.
• To create a database of GO term predictions for proteomes across multiple organisms within the UniProt Knowledgebase (UniProtKB).

Main Methods:

• UniGOPred predicts 514 molecular function (MF), 2909 biological process (BP), and 438 cellular component (CC) GO terms for protein sequences.
• The system was evaluated on CAFA2 challenge data and compared against BLAST classifiers and UniProtKB/TrEMBL annotations.
• The study also explored UniGOPred's capability to predict negatively associated GO terms.

Main Results:

• UniGOPred ranked among the top 10 methods for molecular function prediction in the CAFA2 challenge.
• Its performance surpassed the baseline BLAST classifier across all GO categories.
• Predictions were validated through case studies involving PTEN and CHD8 protein variants.

Conclusions:

• UniGOPred is a powerful, open-access tool for comprehensive protein functional annotation, covering a wide spectrum of GO terms.
• The system demonstrates high accuracy and outperforms existing methods, offering valuable predictions including negative functional associations.
• UniGOPred facilitates deeper understanding of protein roles and variations in biological contexts.