The electron density score for individual atoms (EDIA) quantifies the electron density fit of each atom in a crystallographically resolved structure. Multiple EDIA values can be combined with the help of the power mean to compute EDIAm, the electron density score for multiple atoms to score a set of atoms such as a ligand, a residue, or an active site.
EDIA is calculated by computing the weighted mean over all relevant grid points in the sphere of interest around the atom. The radius of the sphere of interest is two times the resolution dependent electron density sphere radius. More details can be found in the EDIA publication. EDIA values above 0.8 mark well supported, EDIA values falling below 0.8 mark atoms with increasing inconsistency with the electron density. Those should be visually inspected.
Substructures such as residues and ligands can be automatically analyzed with EDIAm. Due to the power mean, an EDIAm below 0.8 has at least one atom with an EDIA below 0.3 or three atoms with EDIAs below 0.4. Substructures with an EDIAm below 0.8 should be visually inspected. Depending on the use case, higher or lower EDIAm values might be tolerable.
1. EDIA or B factor – which one should I use?
The B factor describes disorder of the atom and its value is optimized in the structure refinement. Due to the constraint optimization, B factors of neighboring atoms are interconnected. In combination with the occupancy they can also hide structure modeling errors in the crystallographic refinement procedure.
EDIA in contrast is a purely atom-based scoring scheme with a predefined acceptance for average disorder. It detects sudden changes in the electron density support of the atoms and reports the type of the change with the EDIA error analysis. As modelers, we prefer EDIA over B factors. A crystallographer can use EDIA to double check B factors and annotate them with additional information.
2. EDIA or RSCC/RSR – which one should I use?
RSCC and RSR both calculate the agreement between observed and calculated electron density. Implementations of both differ in the calculation of the element’s electron density radius and the factors used in calculating electron density. While RSCC and RSR are sensitive to the electron density shape used in the comparison, EDIA counters the influence of shape in explicitly taking the relative amount of density in the atoms vicinity into account. Thus, EDIA can balance a fuzzy density border with the help of high electron density intensity and vice versa. In our opinion, EDIA with its well-defined equation and parameters should be used over RSCC and RSR.
3. EDIA or RSZD+RSZO – which one should I use?
RSZD is computed based on the difference density map which is in turn dependent on calculated electron density. If the calculated density depends on B factors and occupancies, it can have an unexpected shape in areas with high B factors or low occupancy. In this area, significant peaks as detected by RSZD cannot occur and are therefore not marked as weakly supported by electron density. Some of these atoms can on the other hand be detected by RSZO which represents the local signal to noise ratio in the difference map.
RSZD and RSZO reflects the crystallographers perspective,showing whether the model is consistent with electron density. EDIA reflects more the viewpoint of the modeler showing whether an individually placed atom is reasonably supported by electron density.
4. Can EDIA values be used for the validation of structure prediction?
The root-mean-standard-deviation (RMSD) is the typical measure for validating structure prediction methods like molecular docking. RMSD uses absolute coordinates and compares against the model. EDIA values can also be computed for structure prediction results comparing a prediction against the experimental outcome, the electron density. In this way EDIA considers the experimental uncertainties visible in the electron density stretch. It describes the closeness to the experiment in the RMSD interval 0 - 2 Angstrom with a higher degree of detail than RMSD.We strongly encourage to supplement RMSD calculation with EDIA in case of method validation.
5. Can EDIA be used for automatically assembling validation data sets of high quality structures?
Yes. EDIA is part of the toolchain in the creation of the Platinum data set. The Platinum data set is a set of over 4,000 high quality protein structures for the validation of conformer generators and other tools.
6. What is the difference between EDIA and EDIAm?
EDIA calculates the electron density support for one atom in the model. EDIAm combines all EDIA scores for a molecular fragment to an indicator score for the electron density support of the fragment. EDIAm drops below 0.8 if at least three atoms have an EDIA score below 0.8.
7. My EDIAm is below 0.8 but there are only two weakly supported atoms. How is this possible?
EDIAm is calculated with the power mean with -2 as exponent to serve as a soft minimum function. If one of the weakly supported atoms has an EDIA value below 0.3, EDIAm drops below 0.8 even though just two atoms are weakly supported. This behavior follows the point of view of a modeler who prefers to flag such structures. A subsequent visual inspection may determine the structure to still be useful depending on the expected usage of the dataset.
8. Can EDIA be used in refinement?
EDIA assists in refinement due to its ability to report actual intensities around a modeled atom. So far, however, EDIA cannot be used in an automatic refinement procedure.
9. Can EDIA discriminate between elements in refinement?
Not automatically but EDIA can give hints in certain cases.
EDIA uses structure factors to calculate the expected electron density radii for each element with a specific charge. Organic elements such as carbon, oxygen and nitrogen have very similar electron density radii. In combination with an electron density grid, differences in radii cannot be detected for the discrimination between those elements.
On the other hand, ions may have vastly different electron density radii. EDIA can detect over-saturated zinc and under-saturated calcium ions with its error analysis. Those together with over-saturated water oxygens and under-saturated chlorine ions should be visually inspected.
EDIA can be calculated for all PDB structures with electron density deposited in the PDBe on our ProteinsPlus server.
The command-line tool EDIAscorer is freely available for academic users for Linux(64 and 32bit) as well as Mac and Windows as part of our NAOMI ChemBio Suite. Non-academic users can get an evaluation license free of charge. No setup steps are required to run EDIAscorer. All feedback is highly appreciated.
People and References
EDIA has been developed by Agnes Meyder, Eva Nittinger and Matthias Rarey at the ZBH - Center for Bioinformatics, University of Hamburg as well as Gudrun Lange and Robert Klein at Bayer AG, Frankfurt, Germany. EDIA for waters was published by Nittinger et al. EDIA and EDIAm were pusblished by Meyder et al.