TripletGO

TripletGO predicts Gene Ontology (GO) terms for protein-coding and non-coding genes by integrating a triplet-network-based expression profiling method with genetic and protein sequence alignments and a Naïve probability model.

Key Features:

• Expression profile similarity (triplet network): Uses transcript expression profiling with a triplet-network-based method to map feature spaces and recognize functional patterns from transcript expressions.
• Genetic sequence alignment: Employs genetic sequence alignment to identify GO terms via comparative genomics, evolutionary conservation, and sequence homology.
• Protein sequence alignment: Uses protein-level alignments to inform functional annotation by comparing amino acid sequences across species.
• Naïve Probability: Applies a basic probabilistic/statistical model to predict potential GO terms based on existing data distributions.
• Neural network integration: Synthesizes outputs from the multiple pipelines via a neural network to produce comprehensive GO term predictions.
• Benchmark evaluation: Validated on 5,754 genes across eight species (human, mouse, Arabidopsis, rat, fly, budding yeast, fission yeast, and nematoda) and 2,433 proteins with expression data from the CAFA3 experiment, demonstrating improved coverage and accuracy versus existing methods.

Scientific Applications:

• GO term prediction for coding and non-coding genes: Annotation of molecular function, biological process, and cellular component GO terms for protein-coding and non-coding genes.
• Cross-species functional inference: Functional annotation leveraging comparative genomics and sequence homology across species including human, mouse, Arabidopsis, rat, fly, budding yeast, fission yeast, and nematoda.
• Benchmarking in CAFA3: Performance assessment and comparison using 2,433 proteins with expression data from the CAFA3 experiment and broader multi-species gene sets.

Methodology:

Combines a triplet-network-based expression profile similarity (feature-space mapping of transcript expression) with genetic sequence alignment, protein sequence alignment, and a Naïve probability model, and integrates pipeline outputs via a neural network.