ProSol-multi: Protein solubility prediction via amino acids multi-level correlation and discriminative distribution

Area of Science:

• Biotechnology
• Bioinformatics
• Computational Biology

Background:

• Protein solubility is crucial for drug development, recombinant protein synthesis, and disease biomarker identification.
• Experimental methods for protein solubility prediction are often time-consuming, costly, and prone to errors.
• Existing AI-based predictors struggle to fully capture discriminative amino acid distributions for accurate solubility prediction.

Purpose of the Study:

• To develop a robust computational predictor for accurate protein solubility prediction.
• To introduce a novel Multi-Level Correlation and Discriminative Distribution Encoding (MLCDE) method for protein sequence analysis.
• To improve the discrimination between soluble and insoluble proteins compared to existing methods.

Main Methods:

• Development of the MLCDE encoder to transform protein sequences into statistical vectors capturing amino acid correlations and distributions.
• Utilizing a Random Forest classifier in conjunction with the MLCDE encoder for protein solubility prediction.
• Evaluation of the MLCDE encoder against 56 existing methods and the ProSol-Multi predictor against 20 existing predictors on benchmark datasets.

Main Results:

• The MLCDE encoder generated non-overlapping clusters for soluble and insoluble protein classes in intrinsic evaluation.
• Machine learning classifiers using the MLCDE encoder outperformed 56 existing encoders in extrinsic evaluation.
• The ProSol-Multi predictor achieved superior performance, outshining 20 existing predictors by an average accuracy of 3% and MCC/AU-ROC of 2% across four datasets.

Conclusions:

• The ProSol-Multi predictor, powered by the MLCDE encoder, offers a significant advancement in computational protein solubility prediction.
• This AI-driven approach provides a more accurate and efficient alternative to experimental methods for selecting candidate proteins.
• The ProSol-Multi web application is available for broader use in research and drug development.