Automated side-chain correction based on electron density

Experimentally determined protein structures stored in the Protein Data Bank (PDB) contain a variety of errors. Thousands of structures are downloaded daily propagating these flaws to the laboratories around the world. This hampers molecular dynamics simulations, electrostatic potential surface computations, and algorithms used in lead design such as in-silico docking and virtual screening. Therefore, highest priority shall be given to the detection and correction of such problems. One common error in structures determined by X-ray crystallography is the incorrect assignment of the terminal side-chain atoms of asparagine (Asn), glutamine (Gln) and histidine (His). In this case the side-chain atoms are fit into an electron density, which appears symmetric resulting in an ambiguous solution consisting of two possible rotamers related by a 180-degree rotation. Current refinement protocols are incapable to pick the right solution automatically. As a consequence, every fourth Asn/Gln/His residue has been deposited incorrectly with the PDB causing hundreds of thousands erroneously annotated atoms, violating basic physico-chemical principles. Over the last years the number of protein structures with available experimental electron density maps has grown immensely and offers a new, previously inaccessible mechanism to remove these errors with unprecedented accuracy: For high resolution structures we demonstrate that the chemical element of ambiguous Asn/Gln/His side-chain atoms can be inferred from the experimental electron density alone and therefore reveals the correct rotamer orientation. This method is introduced as a complementary approach to our previously published method, NQ-Flipper. We show by selected examples that this approach reliably identifies incorrect rotamers missed by existing methods and demonstrate that even difficult cases can readily be solved without ambiguity. Based on these examples we suggest several possible computational approaches to tackle this problem. The Joint Center for Structural Genomics (JCSG) located in San Diego (USA) has just submitted its 1000-th protein structure to PDB and is especially committed to provide high quality models derived in a semi- automated, well documented, and reproducible way. JCSG therefore offers an ideal environment for implementing, incorporating, evaluating, establishing and propagating the proposed method, for receiving premium quality training on high throughput X-ray protein structure determination and as a central hub for building networks in the structure determination community. As such, the so-gained knowledge and networks will be of valuable benefit when reentering the Austrian research community.