Comparison and integration of computational methods for deleterious synonymous mutation prediction.

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Synonymous mutations, despite not altering amino acids, can impact gene function by affecting mRNA structure and stability.
• Next-generation sequencing has revealed numerous synonymous mutations in the human genome, necessitating accurate prediction methods.
• Existing computational models for predicting deleterious synonymous mutations require systematic evaluation and improvement.

Purpose of the Study:

• To systematically compare and evaluate the performance of 10 computational methods for predicting deleterious synonymous mutations.
• To assess the robustness and scalability of these methods using curated independent test datasets.
• To develop an improved ensemble model for more accurate prediction of deleterious synonymous mutations.

Main Methods:

• Systematic comparison of 10 computational methods based on algorithms, features, performance, and usability.
• Construction and utilization of two independent, curated test datasets for rigorous evaluation.
• Development of an ensemble model (PrDSM) by averaging predictions from the top three accurate methods.

Main Results:

• The ensemble model PrDSM demonstrated superior performance compared to individual reviewed tools in predicting deleterious synonymous mutations.
• Benchmark tests confirmed the effectiveness of the ensemble approach for this prediction task.
• An accessible online predictor, PrDSM, was developed based on the ensemble model.

Conclusions:

• The developed ensemble model PrDSM offers a significant advancement in the prediction of deleterious synonymous mutations.
• The comprehensive survey and proposed modeling strategy provide valuable guidance for future research in this field.
• The PrDSM online tool is available to facilitate research and applications in understanding synonymous mutation impacts.