CrowdGO

CrowdGO integrates predictions from multiple computational methods using a wisdom-of-the-crowd consensus approach to generate Gene Ontology (GO) annotations for gene function prediction.

Key Features:

• Integration of Multiple Predictions: Aggregates input predictions from deep learning-based, sequence similarity-based, and protein domain-based algorithms to improve precision and recall in GO term prediction.
• Semantic Similarity Utilization: Employs semantic similarity measures between GO terms to assess consensus among different predictive algorithms.
• Information Content Analysis: Considers the information content of each GO term to refine and prioritize functional annotations.
• Machine Learning Enhancement: Uses a Support Vector Machine (SVM) model to further refine consensus predictions and resolve conflicting annotations.

Scientific Applications:

• Enhanced Functional Annotation: Integrates multiple prediction sources to produce more comprehensive and precise gene functional annotations.
• Benchmarking Performance: Produces consensus predictions that match or exceed the best-performing individual methods in community benchmarks.

Methodology:

Processes outputs from multiple prediction methods (deep learning-based, sequence similarity-based, protein domain-based), computes GO term semantic similarities and information content, re-evaluates gene-term annotations through a consensus approach, and applies a Support Vector Machine (SVM) to refine and produce high-confidence annotations.