Liquid-Liquid Phase Separation Prediction of Proteins in Salt Solution by Deep Neural Network

Area of Science:

• Biophysics
• Computational Biology
• Biochemistry

Background:

• Liquid-liquid phase separation (LLPS) is crucial for forming membrane-free organelles and implicated in disease.
• Protein phase behavior, including phase boundaries and cloud point temperature, is key to understanding LLPS.
• Theoretical and experimental studies have extensively explored protein phase behavior.

Purpose of the Study:

• To develop and apply a deep neural network model for predicting protein liquid-liquid phase separation (LLPS).
• To analyze the phase behavior of lysozyme and bovine serum albumin (BSA) using regression and classification neural networks.
• To validate the model's predictive accuracy for protein solubility, cloud point temperature, and reentrant phase behavior.

Main Methods:

• Utilized regression and classification neural networks to model protein phase behavior.
• Predicted lysozyme solubility and cloud point temperature in NaCl solutions.
• Classified the reentrant phase behavior of BSA in YCl3 and DDAO solutions.
• Experimentally validated model predictions at selected points.

Main Results:

• The neural network model accurately predicted lysozyme solubility and cloud point temperature within specified ranges.
• The model successfully predicted the reentrant phase behavior of BSA under varying conditions.
• Experimental validation confirmed the high accuracy of the deep neural network predictions.
• The model demonstrated capability in both qualitative and quantitative analysis of LLPS.

Conclusions:

• Deep neural networks are effective tools for analyzing protein liquid-liquid phase separation (LLPS).
• The developed model offers accurate predictions for protein phase behavior, aiding in LLPS research.
• This approach has significant potential for advancing the understanding of LLPS in biological systems and disease.