AbsoluRATE

AbsoluRATE predicts absolute aggregation rates of native proteins and peptides using a support vector machine (SVM) regression model to quantify aggregation kinetics under near-physiological conditions and relate sequence attributes to mechanistic behavior.

Key Features:

• SVM-based regression model: AbsoluRATE employs a support vector machine (SVM) regression trained on experimentally determined absolute protein aggregation rates to predict apparent rate constants (kapp, h−1).
• Performance metrics: Model evaluation by leave-one-out cross-validation on 82 non-redundant proteins and peptides produced a Pearson correlation of 0.72 and a mean absolute error (MAE) of 0.91 in natural log scale for kapp.
• Experimental-condition integration: Predictions incorporate experimental variables including temperature, pH, ionic strength, and protein concentration.
• Sequence-based features: The model integrates sequence-derived properties such as inherent aggregation propensity and the presence of aggregation-prone regions (APRs) and gatekeeping residues.
• Mechanism–kinetics coupling: Analysis links common sequence attributes to both aggregation mechanism and kinetics, providing simultaneous mechanistic and kinetic insights.

Scientific Applications:

• Disease progression prediction: Quantitative aggregation-rate predictions can inform analyses of diseases associated with amyloidosis.
• Biotherapeutic storage optimization: Predicted accumulation rates under specified conditions can guide formulation and storage strategies for biotherapeutics.
• Nano-biomaterial engineering: Predicted aggregation kinetics can be used to design and tune nano-biomaterials with desired aggregation behaviors for commercial applications.

Methodology:

An SVM regression model was trained on experimentally determined absolute aggregation rates using features from experimental conditions (temperature, pH, ionic strength, protein concentration) and sequence-based properties (aggregation propensity, APRs, gatekeeping residues), and evaluated by leave-one-out cross-validation on 82 non-redundant proteins and peptides reporting correlation 0.72 and MAE 0.91 (natural log scale) for kapp (h−1).