K-Fold

K-Fold predicts the kinetic order (two-state versus multi-state) and estimates the logarithm of folding rates of proteins from atomic structures to characterize protein folding mechanisms.

Key Features:

• Kinetic Order Prediction: Predicts whether a protein follows two-state or multi-state folding kinetics by analyzing atomic-level structural information.
• Folding Rate Estimation: Estimates the logarithm of the folding rate with a reported correlation of 0.74 to experimental data and a standard error of 1.2.
• Classification Accuracy: Achieves 81% accuracy in classifying folding mechanisms on its training dataset of 63 proteins.
• Machine Learning Model: Implements a support vector machine (SVM) trained on an experimental dataset comprising 63 proteins with well-characterized three-dimensional structures and known folding mechanisms.
• Input Data: Operates on atomic-level three-dimensional protein structures as input for feature extraction and prediction.

Scientific Applications:

• Structural Biology: Elucidating protein folding pathways and classifying folding mechanisms from atomic structures.
• Drug Design: Informing assessments of protein stability and conformational dynamics relevant to therapeutic targeting.
• Biotechnology: Predicting folding behavior of engineered proteins to anticipate stability and kinetic properties in applied contexts.

Methodology:

K-Fold uses a support vector machine (SVM) trained on an experimental dataset of 63 proteins with known three-dimensional structures and folding mechanisms, using features derived from atomic structures.