DeepPhoPred: Accurate Deep Learning Model to Predict Microbial Phosphorylation.

Area of Science:

• Biochemistry
• Computational Biology
• Bioinformatics

Background:

• Phosphorylation is a critical posttranslational protein modification regulating essential cellular processes.
• Predicting microbial phosphorylation sites aids in understanding pathogenesis and developing antimicrobial agents.
• Experimental prediction methods are time-consuming and expensive, necessitating computational solutions.

Purpose of the Study:

• To introduce DeepPhoPred, a novel deep learning tool for predicting microbial phospho-serine (pS), phospho-threonine (pT), and phospho-tyrosine (pY) sites.
• To provide a low-cost, high-speed computational alternative for phosphorylation site prediction in microbes.

Main Methods:

• DeepPhoPred utilizes a two-headed convolutional neural network architecture.
• The model incorporates squeeze and excitation blocks for feature learning.
• It integrates peptide structural and evolutionary information for prediction.

Main Results:

• DeepPhoPred demonstrates superior performance compared to existing microbial phosphorylation site predictors.
• The deep-learning architecture effectively learns significant features for accurate prediction.
• Empirical results validate the tool's high efficiency and accuracy.

Conclusions:

• DeepPhoPred offers a significant advancement in computational prediction of microbial phosphorylation sites.
• The tool's performance and efficiency make it valuable for research in microbial pathogenesis and drug development.
• DeepPhoPred, its code, and datasets are publicly available for broader scientific use.