SAAFEC-SEQ

SAAFEC-SEQ predicts changes in protein thermodynamic stability by estimating folding free energy (ΔΔG) differences caused by single point (missense) amino acid substitutions using sequence information.

Key Features:

• Sequence-based: Operates from amino-acid sequence without requiring three-dimensional structural data.
• ΔΔG prediction: Predicts folding free energy (ΔΔG) changes resulting from single point (missense) amino acid substitutions.
• Feature encoding: Encodes physicochemical properties, sequence-derived features, and evolutionary information as model inputs.
• Machine learning model: Employs a gradient boosting decision tree regressor to map encoded features to stability changes.
• Genome-scale applicability: Enables genome-scale investigations where structural information is limited.
• Benchmark performance: Demonstrated improved performance over other sequence-based methods measured by Pearson correlation coefficient and root-mean-squared-error across multiple independent datasets.

Scientific Applications:

• Protein engineering: Guides selection and design of mutations to increase or modulate protein stability.
• Variant interpretation: Facilitates assessment of disease-associated missense variants by evaluating their impact on protein stability.
• Genome-scale screens: Supports large-scale evaluation of mutation effects when three-dimensional structures are unavailable.

Methodology:

Encodes physicochemical properties, sequence features, and evolutionary information and uses a gradient boosting decision tree regressor to predict folding free energy (ΔΔG) changes from sequence without requiring three-dimensional structure.