Precise prediction of phase-separation key residues by machine learning.

Area of Science:

• Biochemistry
• Molecular Biology
• Genetics

Background:

• Intracellular phase separation is vital for cellular processes like transcription and cell fate determination.
• Identifying key residues that drive protein phase separation has been challenging.
• Dysregulation of phase separation is implicated in various diseases.

Purpose of the Study:

• To develop a machine learning tool, PSPHunter, for accurate prediction of key residues influencing protein phase separation.
• To validate the functional significance of predicted key residues in biological systems.
• To analyze the enrichment of specific amino acids within key residues.

Main Methods:

• Development of a machine learning model (PSPHunter) to predict key residues.
• In vivo and in vitro experimental validation of PSPHunter predictions using GATA3.
• Comprehensive bioinformatic analysis of residue enrichment and disease-associated proteins.

Main Results:

• PSPHunter accurately predicts key residues that govern protein phase separation.
• Truncating six key residues in GATA3 significantly disrupted phase separation, affecting tumor cell migration and growth.
• Glycine and its motifs were found to be enriched in key residues.
• PSPHunter identified disease-associated residues in nearly 80% of studied proteins, often involving glycine and proline.

Conclusions:

• PSPHunter is a powerful tool for identifying critical residues in phase-separating proteins.
• Understanding these key residues provides crucial insights into the mechanisms of transcriptional control, cell fate, and disease development.
• The findings highlight the importance of specific residues, such as glycine and proline, in phase separation and disease pathology.