The final Gibbs free energy change (ΔΔG) in kcal/mol is predicted using more than 70 sequential, molecular and energetic features with the Stochastic Gradient Boosting of Decision Trees algorithm.
The ELASPIC2 (EL2) score is obtained using features generated by pretrained ProteinSolver and ProtBert neural networks with the Stochastic Gradient Boosting of Decision Trees algorithm and a ranking objective function.