Prediction of protein solubility based on sequence physicochemical patterns and distributed representation...

Area of Science:

• Biochemistry
• Bioinformatics
• Computational Biology

Background:

• Protein solubility is essential for heterologous expression and functional studies, but inclusion body formation hinders progress.
• Current protein solubility prediction models are insufficient given the rapid growth of protein sequence data.
• Developing accurate predictors is vital for prioritizing soluble protein targets and reducing experimental costs.

Purpose of the Study:

• To develop a novel, highly accurate tool for predicting protein solubility.
• To improve upon existing protein solubility prediction models.

Main Methods:

• Utilized a long-short-term memory (LSTM) network architecture.
• Incorporated hybrid features including physicochemical patterns and amino acid distributed representations.
• Trained and evaluated the model on protein solubility prediction tasks.

Main Results:

• The developed tool, DeepSoluE, demonstrated superior accuracy and balanced performance compared to existing methods.
• Identified key features significantly impacting model performance and their interactions.
• Achieved more accurate and balanced protein solubility predictions.

Conclusions:

• DeepSoluE is an effective bioinformatics tool for predicting protein solubility in E. coli.
• The tool can prescreen potentially soluble protein targets, significantly reducing the costs of wet-experimental studies.
• A publicly accessible webserver for DeepSoluE is available for free use.