SADA

SADA is a software tool to address the challenge of improving the accuracy of full-chain protein structure modeling. While AlphaFold2 has made significant progress in predicting the correct folds for single-domain proteins, full-chain modeling tends to have lower accuracy on average and requires substantial computational resources. 'SADA' explores the potential for enhancing full-chain modeling accuracy by employing deep learning-assisted domain assembly.
In SADA, a multi-domain protein structure database is created for detecting full-chain analogues using individual domain models. The domain assembly process begins with an initial model generated from the analogues. A two-stage differential evolution algorithm, guided by an energy function incorporating inter-residue distance potentials predicted by deep learning, is employed to simulate the domain assembly and generate the full-chain model.