Computational methods for signal peptide prediction: From statistical models to deep learning.

Area of Science:

• Molecular Biology
• Bioinformatics

Background:

• Signal peptides are N-terminal amino acid sequences crucial for protein localization and transport.
• Experimental identification methods are laborious and costly, necessitating computational approaches.

Purpose of the Study:

• To systematically review and summarize computational methods for signal peptide prediction.
• To analyze the evolution of these methods and their framework designs.
• To identify limitations and discuss future opportunities in computational signal peptide identification.

Main Methods:

• Review of computational approaches developed over the past two decades.
• Comparison of prediction accuracy and methodological frameworks.
• Analysis of limitations and emerging trends in the field.

Main Results:

• Computational methods have progressed from statistical and rule-based algorithms to advanced deep learning techniques.
• Continuous improvement in prediction accuracy has been observed.
• Various computational frameworks have been proposed, each with distinct designs and outcomes.

Conclusions:

• Computational tools are essential for efficient signal peptide identification.
• Future development should focus on unified evaluation, biological interpretation, and generative modeling.
• Advancements aim to enhance the accuracy and interpretability of signal peptide prediction frameworks.