Surveying alignment-free features for Ortholog detection in related yeast proteomes by using supervised big data...

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Ortholog detection is crucial for functional genomics, with previous work focusing on supervised, alignment-based methods in big data platforms.
• Existing supervised models utilized pairwise protein features and low ortholog pair ratios, tested on yeast proteomes.
• Prior approaches relied heavily on alignment-based features and were evaluated on a single test set.

Purpose of the Study:

• To assess the impact of alignment-free features on supervised ortholog detection models.
• To evaluate these models within the Spark big data platform for pairwise ortholog detection in yeast proteomes.
• To compare the performance of alignment-free features against alignment-based features.

Main Methods:

• Implementation of supervised models (Random Forest, Decision Trees) in the Spark big data platform.
• Utilization of both alignment-based and alignment-free pairwise protein features.
• Application of oversampling and undersampling techniques for imbalance management.

Main Results:

• Spark Random Forest and Decision Trees showed high classification performance for ortholog detection in yeast proteomes.
• No significant difference in performance was observed between using only alignment-based features versus combining them with alignment-free features.
• A 98.71% success rate in the 'twilight zone' was achieved with combined features and imbalance management in Spark Decision Trees for a yeast proteome pair with whole genome duplication.

Conclusions:

• Incorporating alignment-free features into supervised big data models did not substantially enhance ortholog detection compared to alignment-based measures alone in yeast.
• The comparable performance of these models to traditional methods suggests potential for alignment-free descriptors.
• Future research should explore additional alignment-free protein pair descriptors for improved ortholog detection.