Improved protein structure refinement guided by deep learning based accuracy estimation

We develop a deep learning framework (DeepAccNet) that estimates per-residue accuracy and residue-residue distance signed error in protein models and uses these predictions to guide Rosetta protein structure refinement. The network uses 3D convolutions to evaluate local atomic environments followed by 2D convolutions to provide their global contexts. The network was trained on approximately 1 million alternative local energy minima for 7,510 different proteins exhibiting a wide diversity of errors, and outperforms other methods that similarly predict the accuracy of protein structure models without template or evolutionary information. Overall accuracy predictions for X-ray and cryoEM structures in the PDB correlate with resolution, and the network should be broadly useful for assessing accuracy of both predicted structure models and experimentally determined structures, and identifying specific regions likely to be in error. Guiding protein structure refinement by incorporation of the accuracy predictions at multiple stages in the Rosetta refinement protocol led to improvements in model quality in 63 out of 73 test cases, illustrating how deep learning can improve search for global energy minima. Significance StatementWe develop a deep learning method to predict the accuracy of protein structure models, and use the method to improve protein structure refinement. Benchmark tests show that both the accuracy prediction method and the protein structure refinement method improve on previously described approaches.