A discriminative method for protein remote homology detection based on N-Gram

Area of Science:

• Bioinformatics
• Computational Biology
• Structural Bioinformatics

Background:

• Protein remote homology detection is crucial for understanding protein function and evolution.
• Existing methods like pairwise comparisons and generative models have limitations in identifying proteins with low sequence similarity.
• Discriminative classifiers offer a promising avenue for improved detection accuracy.

Purpose of the Study:

• To develop and evaluate a novel discriminative classification method for protein remote homology detection.
• To improve the accuracy of detecting structural homology in proteins with low sequence similarity.
• To assess the performance of the proposed method against established techniques.

Main Methods:

• Utilized a discriminative classification approach incorporating N-Grams for feature extraction.
• Employed a random forest algorithm for classifying protein data sets.
• Applied a heuristic method to determine score thresholds for data set division.

Main Results:

• Achieved a 6% improvement in the receiver operating characteristic (ROC) compared to well-known methods on the SCOP 1.53 dataset.
• The heuristic method yielded a precision of 0.5647 and a recall rate of 0.8647 for positive examples.
• Demonstrated superior performance in protein remote homology detection.

Conclusions:

• The N-Grams-based discriminative classification method is effective for protein remote homology detection.
• The proposed approach offers a significant advancement over existing methods, particularly for proteins with low sequence similarity.
• Further investigation into recall and precision metrics for positive examples in this domain is warranted.