Machine Learning-Guided Engineering of Protein Phase Separation Properties in Immune Regulation

Area of Science:

• Biochemistry
• Molecular Biology
• Computational Biology

Background:

• Phase separation (PS) is crucial for cellular compartmentalization and the formation of membraneless organelles.
• Engineering protein PS properties is vital for cellular process regulation but challenging via single amino acid changes.

Purpose of the Study:

• To develop a machine learning tool, PScalpel, for predicting and guiding protein PS engineering.
• To demonstrate PScalpel's efficacy in modulating protein PS ability and cellular functions.

Main Methods:

• Developed PScalpel, a machine learning tool utilizing transfer learning for protein PS prediction.
• Applied PScalpel to engineer TDP43 for enhanced prediction accuracy.
• Engineered the cGAS protein by altering a single amino acid to modulate its PS ability and immune function.

Main Results:

• PScalpel significantly improved prediction accuracy for TDP43, a neurodegenerative disease-associated protein.
• Single amino acid alteration in cGAS successfully modulated its PS ability, optimizing immune function in engineered macrophages.
• Transcriptomic analysis confirmed altered macrophage immune function due to cGAS PS modulation.

Conclusions:

• PScalpel is an effective tool for targeted protein PS engineering.
• Precise biomolecular engineering can be achieved by manipulating protein PS ability.
• This approach offers new methods for targeted molecular and cellular modifications.