Protein-sol

Protein-sol predicts protein solubility and related biophysical properties by applying machine learning to antibody biophysical datasets to model charge and hydrophobicity contributions across 12 distinct biophysical measurements.

Key Features:

• Machine Learning Models: Derives predictive models for 12 distinct biophysical properties associated with protein solubility using machine learning trained on large datasets of antibody biophysical properties.
• Physicochemical Analysis: Analyzes physicochemical determinants, showing that predictive models cluster primarily according to charge and hydrophobicity and linking charge-related predictions to cross-interaction measurements and hydrophobicity to self-interaction methods.
• Hydrophobic Interaction Chromatography: Integrates charge and hydrophobicity data to interpret variations in hydrophobic interaction chromatography (HIC) and protein behavior under different conditions.
• Extensibility to Diverse Proteins: Initially developed from differences observed in antibody variable loops and designed for extension to more diverse protein sets.

Scientific Applications:

• Protein solubility prediction: Predicts solubility and aggregation propensity to inform design of proteins with enhanced stability and reduced aggregation potential.
• Synthetic biology and protein engineering: Supports engineering of novel proteins with desired solubility characteristics by providing predictive insights into charge and hydrophobicity effects.

Methodology:

Applies machine learning algorithms to large datasets of antibody biophysical properties, focusing on charge and hydrophobicity as primary determinants and using cross-interaction and self-interaction measurements to generate predictive models for 12 biophysical properties.