GeFF

GeFF predicts Gene Ontology (GO) term annotations for genes across multiple organisms by applying cross-organism ensemble learning to data from the Entrez Gene database.

Key Features:

• Cross-Organism Ensemble Learning: Predicts new GO annotations for a target organism by leveraging annotations from evolutionarily related organisms using ensemble methods.
• Supervised Machine Learning: Trains models on labeled annotation data to infer gene-function relationships.
• Annotation Perturbation Technique: Creates perturbed training sets by randomly deleting a fraction of known annotations to enable models to reconstruct original annotations and discover novel ones.
• Customizable Ensemble Approaches: Provides alternative ensemble strategies to balance the number of predicted annotations against precision.
• Entrez Gene Database Integration: Leverages extensive annotation data from the Entrez Gene database as input for prediction models.

Scientific Applications:

• Accelerating Annotation Curation: Prioritizes novel predicted GO annotations for experimental validation and curation workflows.
• Complementing Existing Annotations: Enhances completeness and reliability of gene function annotations in genomics and bioinformatics studies.

Methodology:

Applies cross-organism ensemble learning and supervised machine learning trained on perturbed annotation sets created by random deletion of known annotations; validated on Homo sapiens, Mus musculus, Bos taurus, Gallus gallus, and Dictyostelium discoideum to assess prediction quantity–precision trade-offs and the ability to predict novel GO annotations without retraining for different target organisms.