Deep Learning-Based Advances In Protein Posttranslational Modification Site and Protein Cleavage Prediction

Area of Science:

• Biochemistry
• Computational Biology
• Proteomics

Background:

• Posttranslational modifications (PTMs) create proteomic diversity in all organisms.
• PTMs include covalent modifications and proteolytic cleavage, crucial for biological understanding.
• Experimental methods like mass spectrometry aid PTM characterization, but computational approaches are needed for over 450 PTM types.

Purpose of the Study:

• To review recent deep learning (DL) based computational approaches for PTM site prediction.
• To review recent advances in DL for proteolytic cleavage prediction, a less-studied PTM.
• To highlight DL architectures, feature encoding, novelty, and tool availability for PTM prediction.

Main Methods:

• Review of recent literature on DL-based PTM prediction.
• Analysis of DL architectures, feature encoding strategies, and novelty of presented approaches.
• Assessment of the availability of computational tools and resources for PTM prediction.

Main Results:

• Numerous DL-based methods have emerged for PTM site prediction.
• Significant progress has been made in DL for proteolytic cleavage prediction.
• Key aspects such as DL architecture and feature encoding are crucial for prediction accuracy.

Conclusions:

• DL approaches are increasingly vital for understanding and predicting PTMs.
• Further research is needed to explore novel DL architectures and applications in PTM prediction.
• The development and accessibility of DL tools will accelerate PTM research.