ProtRe-CN

ProtRe-CN detects protein remote homology by integrating classification methods and network methods within a Learning to Rank framework to improve identification of remote homologous proteins for downstream protein structure and function prediction.

Key Features:

• Learning to Rank integration: Combines classification methods with network methods through a Learning to Rank framework to produce a unified ranking of candidate homologs.
• Classification methods: Incorporates heterogeneous classification approaches to contribute scores used in the ranking.
• Network methods: Incorporates network-based methods to capture relationship information among proteins for ranking.
• False positive correction: Fuses classification and network outputs to correct false positives within the ranking list and reduce predictive bias.
• Benchmark validation: Experimental evaluations on benchmark and independent datasets demonstrate improved accuracy and reliability compared with state-of-the-art predictors.

Scientific Applications:

• Remote homology detection: Identification of remote homologous proteins for homology-based analyses.
• Protein structure and function prediction: Supports downstream prediction of protein structure and function based on detected homologs.

Methodology:

Combines classification methods and network methods via a Learning to Rank framework, fuses their outputs to correct false positives in the ranking list, and evaluates performance on benchmark and independent datasets.