A machine learning model predicts protein stability of annotated and alternate protein isoforms

Area of Science:

• Molecular Biology
• Bioinformatics
• Proteomics

Background:

• Protein stability is crucial for cellular homeostasis.
• Predicting protein stability from amino acid sequences is challenging.
• Existing knowledge of protein degradation machinery is incomplete.

Purpose of the Study:

• To develop a machine learning model for predicting protein stability from N-terminal sequences.
• To identify sequence motifs that regulate protein stability.
• To investigate the stability of alternative translational isoforms.

Main Methods:

• Machine learning model development using N-terminal amino acid sequences.
• Experimental validation of model predictions.
• Analysis of alternative translational isoforms.

Main Results:

• A novel machine learning model accurately predicts protein stability.
• Identification of known and novel sequence motifs governing protein stability.
• Discovery of N-terminal isoforms with significantly different stabilities.

Conclusions:

• The developed model is a valuable tool for protein stability prediction and design.
• N-terminal sequence variations can drastically alter protein stability and function.
• This work advances understanding of protein regulation and isoform diversity.