ProTstab

ProTstab predicts protein melting temperature (Tm) and cellular protein stability from amino acid sequence to support analysis of protein behavior.

Key Features:

• Machine Learning-Based Prediction: Uses Gradient Boosting of Regression Trees (GBRT) to analyze amino acid sequence data and predict protein stability.
• High Performance Metrics: Reports Pearson's correlation coefficients of 0.793 (10-fold cross-validation) and 0.763 (independent blind test) with mean absolute errors of 0.024 (cross-validation) and 0.036 (blind test).
• Large-Scale Application: Enables proteome-scale stability predictions across extensive datasets.
• Superiority Over Existing Methods: Outperforms previously published methods in predicting protein stabilities.

Scientific Applications:

• Protein Engineering: Predicts stability changes to guide design of proteins with increased or decreased stability.
• Disease Research: Assesses how variations affect protein stability to inform studies of disease mechanisms linked to altered stability.
• Protein Function Analysis: Correlates predicted stabilities with protein chain lengths of isoforms and subcellular localizations for functional genomics investigations.

Methodology:

Training data were generated by limited proteolysis combined with mass spectrometry, and the model was trained on amino acid sequence information using Gradient Boosting of Regression Trees (GBRT).