Predicting deleterious non-synonymous single nucleotide polymorphisms in signal peptides based on hybrid sequence...

Area of Science:

• Bioinformatics
• Molecular Biology
• Computational Biology

Background:

• Signal peptides are essential for protein localization and secretion.
• Mutations in signal peptides, known as non-synonymous single nucleotide polymorphisms (nsSNPs), can disrupt protein function.
• Accurate prediction of deleterious nsSNPs is vital for understanding disease mechanisms.

Purpose of the Study:

• To develop and validate a computational method for predicting deleterious nsSNPs in signal peptides.
• To improve the accuracy of nsSNP prediction by integrating multiple feature types.
• To provide a superior alternative to existing nsSNP prediction methods.

Main Methods:

• Utilized a random forest (RF) algorithm incorporating Position Specific Scoring Matrix (PSSM) profiles, SignalP scores, and physicochemical properties.
• Optimized feature selection using the Maximum Relevance Minimum Redundancy (mRMR) method.
• Employed a cost matrix to address data imbalance challenges in nsSNP prediction.

Main Results:

• Achieved an overall accuracy of 84.5% and an Area Under the ROC Curve (AUC) of 0.822 using a Jackknife test with 10 optimized features.
• Demonstrated superior performance compared to existing R-score and D-score-based prediction methods on the same dataset.
• The developed method effectively predicts deleterious nsSNPs in signal peptides.

Conclusions:

• The proposed computational method offers a robust and accurate approach for identifying deleterious nsSNPs in signal peptides.
• This tool can aid researchers in understanding the functional consequences of signal peptide mutations.
• The findings highlight the potential of integrating diverse features for enhanced nsSNP prediction accuracy.