User guide to the wwPDB X-ray validation reports

 

Last updated: 03/25/14

 

1. Overall quality at a glance

2. Entry composition

3. Residue-property plots

4. Data and refinement statistics

5. Model quality

5.1 Standard geometry

5.2 Close contacts

5.3 Torsion angles

5.3.1 Protein backbone

5.3.2 Protein sidechains

5.3.3 RNA

5.4, 5.5, 5.6, 5.7 Non-standard residues in protein, DNA, RNA chains, Carbohydrates, Ligand geometry, Other polymers

6. Fit of model and data

6.1 Protein, DNA and RNA chains

6.2, 6.3, 6.4, 6.5 Non-standard residues in protein, DNA and RNA chains, Carbohydrates, Ligands, Other polymers

References

 

 

 

The new style wwPDB X-ray validation reports are prepared according to the recommendations of the wwPDB X-ray Validation Task Force (VTF; Read et al., 2011). The report summarises the quality of the structure and highlights specific concerns by considering the atomic model, the diffraction data and the fit between the atomic model and the diffraction data.

 

The title page shows some information about the entry deposition as well as the names and version numbers of the software tools and reference information used to produce the report.

1. Overall quality at a glance

This section provides a succinct “executive” summary of key quality indicators. If there should be serious issues with a structure, this would usually be evident from this summary.

 

The metrics shown in the “slider” graphic (see example below) compare several important global quality indicators for this structure with those of previously deposited PDB entries. The comparison is carried out by calculation of the percentile rank, i.e. the percentage of entries that are equal or poorer than this structure in terms of a quality indicator. The global percentile ranks (black vertical boxes) are calculated with respect to all X-ray structures available in the PDB archive prior to 2011. The resolution-specific percentile ranks (white vertical boxes) are calculated with respect to a subset of X-ray entries in the same subset of the PDB archive, but only considering entries with comparable resolution to this entry. In general, one would of course like all sliders to lie far to the right in the blue areas (especially for recently determined structures, and especially the resolution-specific sliders).

 

 

 

Note that as a non-expert you neither need to know what the various quality criteria measure nor whether the values for an entry are unusual or not. However, for increased understanding, below is a brief description of these key global quality indicators. For more information see Read et al. (2011) or the review by Kleywegt (2000).

 

      Rfree: Rfree is a measure of the fit of the model to a small subset of the experimental data, which was not used in model refinement (Brunger, 1992). It is calculated by DCC (Yang et al.).

      Clashscore: This score is derived from the number of pairs of atoms in the model that are unusually close to each other. It is calculated by MolProbity (Chen et al., 2010) and expressed as the number or such clashes per thousand atoms.

      Ramachandran outliers: A residue is considered to be a Ramachandran outlier if the combination of its phi and psi torsion angles is unusual, as assessed by MolProbity (Chen et al., 2010). The Ramachandran outlier score for an entry is calculated as the percentage of Ramachandran outliers with respect to the total number of residues in the entry for which the outlier assessment is available.

      Sidechain outliers: Protein sidechains mostly adopt certain (combinations of) preferred torsion angle values (called rotamers or rotameric conformers), much like their backbone torsion angles (as assessed in the Ramachandran analysis). MolProbity considers the sidechain conformation of a residue to be an outlier if its set of torsion angles is not similar to any preferred combination. The sidechain outlier score is calculated as the percentage of residues with an unusual sidechain conformation with respect to the total number of residues for which the assessment is available.

      RSRZ outliers: The real-space R-value (RSR) is a measure of the quality of fit between a part of an atomic model (in this case, one residue) and the data in real space (Jones et al., 1991). The RSR Z-score (RSRZ) is a normalisation of RSR specific to a residue type and a resolution bin (Kleywegt et al., 2004). RSRZ is calculated only for standard amino acids and nucleotides in protein, DNA and RNA chains. A residue is considered an RSRZ outlier if its RSRZ value is greater than 2. The RSRZ outlier score as shown in the slider graph is calculated as the percentage RSRZ outliers with respect to the total number of residues for which RSRZ was computed. This is calculated by the EDS (Electron-Density Server) component of the validation pipeline which is a re-implementation of the software used by the Uppsala EDS server.

      RNA backbone: Like the protein backbone and sidechains, the RNA backbone also adopts certain sets of preferred torsion angle values. Based on statistical analysis of RNA chains in the PDB, MolProbity (Chen et al., 2010) assigns a score per nucleotide for the quality of its backbone. This metric is calculated as the average score of all nucleotides in the entry.

 

The table below the slider graph (see example above) shows the number of entries upon which the percentile rank calculations are based, and the resolution ranges in which they lie. (The resolution-specific ranks are calculated using the smallest bin width that includes at least 1,000 entries.)

 

 

The next table shows a graphical summary of the quality of all polymeric chains. There may be green, yellow, orange and red portions in the lower bar for each chain, indicating the fraction of residues that contain outliers for 0, 1, 2, >=3 model-only validation criteria. Any grey portion indicates unmodeled residues. If electron density outliers were present, there is an additional red bar above the lower bar, indicating the fraction of residues that are RSRZ outliers.

 

 

The last table (see example above), when present, lists compounds that were perhaps not modelled adequately. The “Geometry” column indicates an issue with an ‘X’ mark if at least 40% of bonds, angles, torsions, chiral centres and rings are outliers. The “Electron density” column indicates an issue with an ‘X’ mark if the compound’s real-space R-value exceeds 0.5 or its Local Ligand Density Fit value (LLDF, described below in section 6.2) is greater than 2.

 

2. Entry composition

 

This section summarises the number of unique molecules that are present in the entry, and how they have been modelled. Each unique molecule and its instances (chain id) are described in a table with the following columns:

      Mol: The identifier of the molecule (for experts: this is the same as the “entity id” in the mmCIF file of the entry).

      Chain: The instance identifier. If there is more than one model present in the entry, the chain is prefixed with a model number.

      Residues: The number of residues in the molecule.

      Atoms: This tabulates the counts of various element types in the molecule.

      ZeroOcc: The number of atoms modelled with zero occupancy in the molecule.

      AltConf: The number of residues that have been modelled with at least one alternative conformation.

      Trace: The number of residues in the molecule that have been modelled with a reduced set of atoms. Protein or nucleic acid chains may be modelled with only one or two atoms (e.g. Cα, Cβ, P, an atom in a sugar ring or nucleobase, etc.). Typically, such cases are observed when the resolution is insufficient to confidently model all atoms.

The molecule and chain identifiers in this section (see example below) are also used in the subsequent tables in the report.

 

 

3. Residue-property plots

 

This section shows summary plots of quality information for protein, RNA and DNA molecules on a per-residue basis.

 

There are two graphics shown for each molecule. The first graphic is the same as that shown in section 1: the green, yellow, orange and red segments indicate the fraction of residues with 0, 1, 2 and 3 or more types of model-only quality criteria with outliers, respectively. The additional red segment above the summary graphic (if present) indicates the fraction of residues that have an unusual fit to the density (RSRZ outliers).

 

The types of model-only quality criteria included in this analysis, and the software used for their calculation are:

      bond length and angle outliers (MolProbity, Chen et al., 2010)

      chirality outliers (Validation-pack, Feng et al.)

      planarity outliers (Validation-pack, Feng et al.)

      close contacts (MolProbity, Chen et al., 2010 and Validation-pack, Feng et al.)

      protein backbone (Ramachandran) outliers (MolProbity, Chen et al., 2010)

      protein sidechain torsion angle outliers (MolProbity, Chen et al., 2010)

      RNA backbone torsion angle outliers (MolProbity, Chen et al., 2010)

      RNA sugar pucker outliers (MolProbity, Chen et al., 2010)

 

The second graphic shows the sequence annotated by these criteria with outliers in model quality and unusual fit to the electron density (see example graphic below). The colour-coding described above is used here too, with a red dot above a residue indicating a poor fit to the electron density (i.e., an RSRZ outlier). Consecutive stretches of residues for which no outliers were detected at all are not shown individually, but indicated by a green connector.

 

In general, the less red, orange and yellow these plots contain, the better. It is important to realise that residues that are outliers on one or more model-validation criteria could be either errors in the model, or reflect genuine features of the structure. Careful analysis of the experimental data (electron density maps) is typically required to make the distinction. Outlier residues that are important for structure or function (e.g., enzymatic residues, interface residues, ligand-binding residues) should be inspected extra carefully (and addressed in a manuscript describing the structure).

 

 

4. Data and refinement statistics

This section provides a summary of the data and refinement statistics for the entry, including the source of the information. The precise definitions of many of the metrics can be found in original publications and crystallography textbooks. The following information is included in the table:

      Space group: This is the symmetry system in which the sample was crystallised to obtain the diffraction data.

      Cell constants: These are the unit-cell dimensions and angles.

      Data completeness: The number of expected diffraction spots is a function of data resolution and the space group. This metric describes the number of recorded reflections  as a percentage of the number expected. It is reported both as provided by the depositor and as calculated by Refmac.

      Rmerge , Rsym: These metrics give an indication of the agreement between multiple intensity measurements.

      <I/σ(I)>: Each reflection has an intensity (I) and an uncertainty in measurement (σ(I)), so I/σ(I) is the signal-to-noise ratio. This ratio decreases at higher resolution. <I/σ(I)> is the mean of individual I/σ(I) values. Value for outer resolution shell is given in parentheses. In case structure factor amplitudes are deposited, Xtriage estimates the intensities first and then calculates this metric. When intensities are available in the deposited file, these are converted to amplitudes and then back to intensity estimate before calculating the metric.

      Resolution: Lower and upper limits of the resolution of the diffraction data, as reported by the depositor (usually, the range of data used in refinement of the model) and as calculated by EDS (the range of data encountered in the deposited reflection file).

      Refinement program: The software that the depositor used for the final crystallographic refinement.

      R, Rfree: These metrics (calculated by DCC) measure the similarity between the observed structure-factor amplitudes and those calculated from the model. Lower values and smaller differences between the two statistics are usually better. Rfree should be higher than R because it is calculated using reflections not used in the refinement.

      Wilson B-value: An estimate of the overall B-value of the structure, calculated from the diffraction data. It serves as an indicator of the degree of order in the crystal and the value is usually not hugely different from the average B-value calculated from the model.

      Anisotropy: The ratio (BmaxBmin) / Bmean where Bmax, Bmin and Bmean are computed from the B-values associated with the principal axes of the anisotropic thermal ellipsoid. This ratio is usually less than 0.5; for only 1% of PDB entries it is more than 1.0 (Read et al., 2011).

      Bulk-solvent parameters: Disordered solvent occupies a significant fraction of the crystal and contributes to low-resolution reflection terms. Bsol and ksol are parameters in a formulation for estimation of bulk solvent contributions to structure factors, and they are estimated during refinement.

      Estimated twinning fraction: Twinning is the phenomenon where at least two domains occur within a single crystal related by mathematical operations called twinning operators. The twinning fraction is estimated using the H-test (Yeates, 1988). The estimated fraction is listed along with the associated twinning operator. Twinning is not possible in all space groups.

      L-test for twinning: This test, based on acentric reflections, defines two metrics < |L| > and < L2 >, whose theoretical values are 0.5 and 0.375 in the untwinned case, and 0.333 and 0.2 in the perfectly twinned case (Padilla & Yeates, 2003).

      Outliers: These are reflections in the diffraction data with very unlikely magnitude for amplitude or intensities. These are identified by Xtriage using its “extreme value statistics” method on the largest normalised intensities (Read, 1999). More than 0.1% outliers in a dataset generally indicates a fundamental problem with the data, unless translational NCS is present. Xtriage’s assessment of possible translational NCS is printed below the table.

      Fo,Fc correlation: The difference between the observed structure factors (Fo) and the calculated structure factors (Fc) measures the correlation between the model and the experimental data. This metric is calculated by Refmac.

      Total number of atoms: This is the total number of atoms modelled in the entry.

      Average B, all atoms: This is the mean B-value calculated over all modelled atoms.

 

5. Model quality

Quality statistics in this section are calculated using standard compilations of covalent geometry parameters (Engh and Huber, 2001; Parkinson et al., 1996), tools in MolProbity (Chen et al., 2010), Validation-pack (Feng et al.) and the wwPDB chemical component dictionary (CCD).

 

5.1 Standard geometry

 

This section describes the quality of the covalent geometry for protein, DNA and RNA molecules in terms of bond lengths, bond angles, chirality and planarity. There are two tables providing a per-molecule summary and four tables that provide information on (some of) the outliers for each criterion (if any; otherwise the table is omitted).

 

Summary table for bond lengths and angles

      Expected bond length and bond angle values (and standard deviations) for standard amino acids and nucleotides are available in a wwPDB compilation (wwPDB, 2012). Dangle (MolProbity) calculates Z-scores of bond length and bond angle values for each residue in the molecule relative to the expected values. (A Z-score is generally defined as the difference between an observed value d an expected or average value, divided by the standard deviations of the latter.)

      The root-mean-square value of the Z-scores (RMSZ) of bond lengths (or angles) is calculated for the whole molecule and also for individual residues. RMSZ scores are expected to lie between 0 and 1. For low-resolution structures, geometry should be tightly restrained and small values are expected. For very high-resolution structures, values approaching 1 may be attained. Values greater than 1 indicate over-fitting of the data. Individual bond lengths or angles with a Z-score greater than 5 or less than -5 merit inspection.

      The bond/angle summary table has the following columns:

      Mol: The molecule identifier

      Chain: The instance identifier.

      Bond lengths: The RMSZ score of all bond lengths is given in the “RMSZ column”.  The number of bond lengths that have a Z-score > 5 or < -5 and the total number of bonds present is given in the “#|Z| >5” column. The percentage of such outliers is listed in parentheses.

      Bond angles: The RMSZ score of all bond angles is given in the “RMSZ” column.  The number of bond angles that have a Z-score > 5 or < -5 and the total number of angles present is given in the “#|Z| >5” column. The percentage of such outliers is listed in parentheses.

 

 

Summary table for chirality and planarity

      Deviations from expected chirality and planarity in the model are calculated by Validation-pack (Feng et al.).

      Chiral centres for all compounds occurring in the PDB are described in the chemical component dictionary. Chirality can be assessed in a number of ways, including calculation of the chiral volume, e.g. for the Cα of amino acids this is 2.6 or -2.6 Ā3 for L or D configurations, respectively. If the sign of the computed volume is incorrect, the handedness is wrong. If the absolute volume is less than 0.7 Ā3, the chiral centre has been modelled as a planar moiety which is very likely to be erroneous. Chirality deviations are summarised per chain.

      Three kinds of potential planarity deviations are assessed:

      Sidechain: Certain groups of atoms in protein sidechains and nucleotide bases are expected to be in the same plane. An atom’s deviation from planarity is calculated by fitting a plane through these atoms and then calculating distance of individual atom from the plane. Expected value of such distances have been pre-calculated from data analysis (wwPDB, 2012). If an atom is modelled to be more than six times farther than the pre-calculated value, the residue is flagged to have a sidechain planarity deviation.

      Peptide: A deviation is flagged if the omega torsion angle of a peptide group differs by more than 30o from the values expected for a proper cis or trans conformation (0Ż and 180Ż, respectively).

      Main chain: The N atom of an amino acid residue is expected to be in the same plane as the Cα, C, and O atoms of the previous residue. If it is out of plane by more than 10o, this is flagged as a planarity deviation.

 

Where outliers exist, up to five for each category are listed in a table. For bond lengths and bond angles, they are the worst outliers. The outlier tables have the following columns in common:

      Mol: The molecule identifier.

      Chain: The instance identifier

      Res: The residue number. Where applicable, an insertion code and alternative conformation identifier are specified as well.

      Type: The residue name.

 

The following columns are specific to the bond length and bond angle outlier tables:

      Atoms: names of atoms involved in the bond or angle.

      Z: The Z-score of the bond length or angle.

      Observed: The observed value of the bond length or angle.

      Ideal: The ideal value of the bond length or angle.

 

The following column is specific to the chirality outliers table:

      Atom: The name of the chiral atom with the unusual deviation.

 

 

The following column is specific to the planarity outliers table:

      Group: The planarity deviation type, i.e. sidechain, main chain or peptide as described above.

 

 

5.2 Close contacts

This section provides details about close contacts (clashes). Two kinds of close contacts are analysed: those that occur between atoms in the same asymmetric unit (ASU) and those that occur between atoms in different ASUs. The latter are called symmetry-related clashes.

 

Close contacts within the ASU are calculated by MolProbity (Chen et al., 2010). Any missing hydrogen atoms are added and all close contacts identified. MolProbity then calculates an all-atom clashscore, which is defined as the number of clashes per 1000 atoms (including hydrogens). Percentile scores of the clashscore are also computed, to allow assessment of how good or bad the clashscore value is.

 

Symmetry-related clashes are identified by Validation-pack (Feng et al.). For each clash, a symmetry and translation code of the ASU of the clashing atom is also reported.

 

Clashes are summarised in a table with the following columns:

      Mol: The molecule identifier

      Chain: The instance identifier

      Non-H: The number of non-hydrogen atoms modelled.

      H(model): The number of hydrogen atoms modelled.

      H(added): The number of hydrogen atoms added by MolProbity. These are not used in the detection of symmetry-related clashes.

      Clashes: The number of clashes in which the atoms in this instance of the molecule are involved.

      Symm-clashes: The number of symmetry-related clashes in which the atoms in this instance of the molecule are involved.

 

 

The “magnitude” of a clash is defined as the difference between the observed interatomic distance and the sum of the van der Waals radii of the atoms involved. Up to five of the worst clashes within the ASU, if indeed there are any, are listed in a table. The table has the following columns:

      Atom-1 and Atom-2: The molecule identifier, instance identifier, residue number, residue name and atom name of both clashing atoms. Where applicable, the chain identifier is prefixed with model number and an alternative conformation identifier is shown as a suffix to the atom name.

      Distance: The interatomic distance between Atom-1 and Atom-2.

      Clash: The magnitude of the clash, as defined above, as provided by Molprobity.

 

Where there are symmetry-related clashes, the (up to) five worst of them are also listed in a table, which is formatted in the same way as the previous table. However, a symmetry and translation identifier for the second atom in the clash is added in square brackets in the Atom-2 column. The first number is the index of the crystallographic symmetry operator; the second number encodes how many unit cell translations in each direction are necessary (“555” is the deposited asymmetric unit, “456” indicates that a fractional translation of (-1 0 +1) is required, etc.).

 

For symmetry-related clashes, the criterion for identifying clashes is different. In this case, the “magnitude” of a clash is defined as 2.2Ā (or 1.6Ā if either atom is a hydrogen) minus the interatomic distance.

 

 

 

5.3 Torsion angles

5.3.1 Protein backbone

 

This section is populated if there are protein molecules present in the entry. The conformation of a protein backbone can be described by a pair of torsion angles (phi, psi) per residue (the remaining torsion angle, omega, is usually 180Ż). Ramachandran plots show the combinations of phi-psi values in a structure and typically compare these to a distribution of commonly observed values in high-resolution crystal structures. MolProbity’s Ramachandran plots are residue-type specific, derived from a high-quality subset of protein X-ray structures and divided into favoured, allowed and outlier regions. Favoured and allowed regions are defined to be the regions that include 98% and 99.95%, respectively, of the residues in the high-quality data (see Chen et al. (2010) for more details).

 

This section contains a summary of analysis of the backbone torsion angles phi and psi by Molprobilty. The summary table contains the following columns:

      Mol: The molecule identifier

      Chain: The instance identifier

      Analysed: The first number here is the number of residues in the chain for which MolProbity output is available. The second number is the total number of residues in the chain. Phi and psi angles cannot be analysed for terminal residues, non-standard residues or for residues with incompletely modelled main chain.

      Favoured, Allowed, Outliers: The number (and percentage) of residues in the favoured, allowed and outlier regions respectively, of the residue-specific phi-psi plots.

      Percentiles: The percentile score based on the percentage of Ramachandran outliers in the chain. These are given relative to the whole archive (first value) and relative to structures of a similar resolution (second value).

 

The outlier table shows up to five randomly chosen outlier residues. It has following columns:

      Mol: The molecule identifier

      Chain: The instance identifier

      Res: The residue number

      Type: The residue name

 

5.3.2 Protein sidechains

 

Protein sidechain conformation can be described by the chi torsion angles. Depending on residue type, these angles adopt certain preferred sets of values (also termed rotamers or rotameric conformers). Based on analysis of high quality X-ray entries in the PDB, MolProbity assesses whether a sidechain is similar to one of the preferred sets of torsion angles, or is an outlier (see Chen et al., 2010 for details). This section is based on MolProbity analysis of sidechains.

 

The summary table summarises of sidechain outliers and has the following columns:

      Mol: The molecule identifier

      Chain: The instance identifier

      Analysed: The first number here is the number of residues in the chain which were analysed by MolProbity. The second number is the total number of residues in the chain. Chi torsion angles cannot be analysed for non-standard residues or for residues with incompletely modelled sidechains.

      Rotameric, Outliers: The number (and percentage) of residues with favoured, and unusual chi torsion angles respectively.

      Percentiles: The absolute and relative percentile scores based on the percentage of sidechain outliers in the chain. These are given relative to the whole archive (first value) and relative to structures of a similar resolution (second value).

 

The outlier table shows up to five randomly chosen outlier residues. It has the following columns:

      Mol: The molecule identifier

      Chain: The instance identifier

      Res: The residue number

      Type: The residue name

 

Side chains of asparagine, glutamine and histidine can sometimes be rotated (“flipped”) to make optimal hydrogen bonds, improving its contacts with its neighbours,  without affecting their fit to the experimental electron density (see Chen et al., 2010 for details).

 

The sidechain flipping table lists up to five sidechains for which being ‘flipped’ improves their contacts with atomic neighbours. It has the following columns:

      Mol: The molecule identifier

      Chain: The instance identifier

      Res: The residue number

      Type: The residue name

5.3.3 RNA

 

This section describes the quality of RNA chains using MolProbity’s analysis of ribose sugar puckers and rotameric nature of ‘suites’ of backbone torsion angles (Richardson et al., 2008, Chen et al., 2010). A suite consists of the torsion angles between the sugars in two RNA nucleotides and is identified by the 3’ nucleotide.

 

The summary table summarises the geometrical quality of an RNA chain using the following columns:

      Mol: The molecule identifier

      Chain: The instance identifier

      Analysed: The first number here is the number of backbone suites for which analysis was carried out, and the latter number is the total number of nucleotides. The former is a smaller number because a suite is not defined at 5’ end, or a suite might be incompletely modelled

      Backbone outliers: The percentage of nucleotide suites in the chain which Molprobitiy identified as an outlier.

      Pucker outliers: The percentage of sugar pucker outliers in the chain which Molprobitiy identified as an outlier. These are nucleotides where the strong correlation between sugar pucker and distance between the glycosidic bond vector and the following phosphate is violated.

 

The backbone and pucker outlier tables provide details of up to five outliers, when applicable, using the following columns:

      Mol: The molecule identifier

      Chain: The instance identifier

      Res: The residue number.

      Type: The residue name.

 

5.4, 5.5, 5.6, 5.7 Non-standard residues in protein, DNA, RNA chains, Carbohydrates, Ligand geometry, Other polymers

 

This section analyses the geometry of:

      non-standard amino acid residues

      non-standard nucleotides within DNA or RNA

      Carbohydrates

      Ligands

      Other polymers

Bond lengths, bond angles, acyclic torsions and isolated rings are assessed (using Mogul; Bruno et al., 2004) by comparison with preferred molecular geometries derived from high-quality, small-molecule structures in the Cambridge Structural Database (CSD). Chirality is assessed by Validation-pack (Feng et al.).

There are two summary tables providing a per-molecule overview and detailed tables that provide information on (some of) the outliers for each criterion (if any; otherwise the table is omitted).

 

Summary table for bond lengths and angles

 

A Z-score is calculated for each bond length and bond angle in the molecule (A Z-score is generally defined as the difference between an observed value and an expected or average value, divided by the standard deviations of the latter.). Individual bond lengths or angles with a Z-score less than -2 or greater than 2 merit inspection.

 

The root-mean-square value of the Z-scores (RMSZ) of bond lengths (or angles) is calculated for the whole molecule. RMSZ scores are expected to lie between 0 and 1. For low-resolution structures, geometry should be tightly restrained and small values are expected. For very high-resolution structures, values approaching 1 may be attained. Values greater than 1 indicate over-fitting of the data.

 

The bond/angle summary table has the following columns:

      Mol: The molecule identifier.

      Type: The residue name.

      Chain: The instance identifier.

      Res: The residue number.

      Link: The identifier(s) of the molecule(s) to which the residue is linked, e.g. by a covalent bond, salt bridge etc.

      The Bond lengths (or angles) column is subdivided into three:

      Counts: This column is sub-divided into 3 sub-columns: the number of bonds (or angles) analysed, the number of bonds (or angles) modelled in the residue and the number of bonds (or angles) defined in the PDB chemical component dictionary. The number of bonds (or angles) analysed may be less than observed due to the absence of comparable fragments in the Cambridge Structural Database.

      RMSZ: The root-mean-square value of the Z-scores (RMSZ) of all bond lengths (or angles) is given in the “RMSZ column”.

      The number of bond lengths or bond angles that have a Z-score of less than -2 or greater than 2 and the total number of bonds / angles present is given in the “#|Z| >2” column. In parentheses the number of outliers within the molecule is listed as a percentage.

 

 

 

Summary table for chirality, torsions and rings

 

For acyclic torsion angles, Mogul provides a distribution of absolute values of torsion angles observed in comparable fragments in the Cambridge Structural Database. This distribution and the observed value are used to find a mean difference and the minimum difference, both of which have to be above 60o for the torsion to be flagged as an outlier.

For isolated rings, Mogul compares the given ring with comparable rings in small molecules structures in the Cambridge Structural Database and calculates an RMSD value based on corresponding constituent torsion angles for each comparable ring. The mean and minimum of these RMSDs both have to be above 60o for the ring to be flagged an outlier.

Note that the criteria used to flag a ring or torsion angle as an outlier are under development. The current criteria are very conservative. They will be refined following analysis of a large test set of ligands.

 

 

The chirality, torsion angles and rings summary table contains the following columns:

      Mol: The molecule identifier.

      Type: The residue name.

      Chain: The instance identifier.

      Res: The residue number.

      Link: One or more molecule identifiers to which the residue is linked, e.g. by a covalent bond, salt bridge etc.

      Chirals: This column lists: the number of chiral outliers in the chain, the number of chiral centers analysed, the number of these observed in coordinates and the number defined in the PDB chemical component dictionary.

      Torsion: This column lists: the number of torsion angle outliers in the chain, the number of torsions analysed, the number of these observed in coordinates and the number defined in the PDB chemical component dictionary.

      Rings: This column lists: the number of ring outliers in the chain, the number of rings analysed, the number of these observed in coordinates and the number defined in the PDB chemical component dictionary.

 

 

Information tables for bond length, bond angle, chirality, torsion angle and ring outliers

 

Where outliers exist, up to five for each category are listed in a table. For outliers of bond lengths and bond angles, they are the worst outliers.

 

The outlier tables have the following columns in common:

      Mol: The molecule identifier.

      Type: The residue name.

      Atom(s): names of atoms involved in the bond, angle, torsion angle, ring, or the name of the chiral atom with the unusual deviation.

      Chain: The instance identifier.

      Res: The residue number.

The following columns are specific to the bond length and bond angle outliers tables:

      Z: The difference between observed and ideal values in terms of standard deviations.

      Observed: The observed value of the bond length or angle.

      Ideal: The ideal value of the bond length or angle.

 

6. Fit of model and data

This section presents analysis of the fit of the molecules in the entry to the experimental data. EDS (Kleywegt et al., 2004) is used to generate experimental electron density using the coordinates in the entry and the experimental data. The experimental electron density is then compared to idealised electron density and the difference reported as a Real Space R-value (RSR). This analysis is performed on a per residue / nucleotide basis. Lower values of RSR indicate a closer match between the experimental and calculated electron density, and thus a better fit of the residue / nucleotide to the experimental data. RSRZ is a derived measure which normalises RSR against residue type and resolution. RSRZ values of greater than 2 indicate an outlier.

RSRZ is defined only for standard amino acids and nucleotides, due to the lack of sufficient data to make statistically significant conclusions for other molecules. Instead RSR is calculated but not normalised.

 

6.1 Protein, DNA and RNA chains

 

This section describes the fit between the experimental electron density and standard residues / nucleotides in protein, DNA and RNA chains. The first table summarises the quality using the following columns:

      Mol: The molecule identifier

      Chain: The instance identifier

      Analysed: This column provides the number of residues in the chain for which RSRZ was analysed over the total number of residues present in the chain. In parentheses this value is expressed as a percentage. 

      <RSRZ>: The mean value of the per-residue RSRZ.

      #RSRZ > 2: This column is separated into three sub columns: the number of residues for which RSRZ is is less than -2 or greater than 2. The next two columns compare the number of residues / nucleotides with RSRZ less that -2 or greater than 2 per number of residues / nucleotides analysed relative to all X-ray structures available in the PDB archive prior to 2011 (2nd column) or a subset with similar resolution to this entry.

      OWAB: The Occupancy-Weighted Average B (OWAB) value per residue.  This value is calculated by multiplying the B factor for each atom in the residue by its occupancy and then averaging this value over all atoms in the residue. The OWAB column is then presented as 4 sub columns providing: the minimum, median, 95th percentile and maximum OWAB value for all residues in the instance being analysed.

      Q<0.9: This is defined as the number of residues in the instance which have an average occupancy of less than 0.9. In parentheses this value is expressed as a percentage of total observed residues / nucleotides in the instance.

 

The residues which have an RSRZ of greater than 2 are listed in a subsequent table. If there are more than five residues which have an RSRZ of greater than 2 then the five with RSRZ scores furthest away from zero are listed.

 

6.2, 6.3, 6.4, 6.5 Non-standard residues in protein, DNA and RNA chains, Carbohydrates, Ligands, Other polymers

 

This section describes the fit between the experimental electron density and:

      non-standard amino acid residues

      non-standard nucleotides within DNA or RNA

      Carbohydrates

      Ligands

      Other polymers

 

for which RSRZ values cannot be calculated due to insufficient instances within the PDB archive for statistically significant conclusions to be drawn (Kleywegt et al. (2004)).

 

The Non-standard residues, Carbohydrates, Ligands and Other polymer tables have the following columns in common:

      Mol: The molecule identifier.

      Type: The name of the molecule.

      Chain: The instance identifier.

      Res: The instance residue number.

      Atoms: The two numbers in this column represent the number of atoms modelled in the instance over the number of atoms defined for the instance type in the chemical component dictionary.

      RSR: The Real Space R-value for the instance.

      LLDF: Local Ligand Density Fit (LLDF) compares the RSR of the non-standard residue to the mean and standard deviation of RSR for the neighbouring polymeric standard  amino acids or nucleotides. Neighbouring polymeric residues are identified as any standard amino acid or nucleotide within 5Ā (Ncont (CCP4)) distance of the non standard residue. The mean and standard deviation of RSR values is then calculated for these neighbouring residues and this is compared to the RSR value of the non-standard residue to calculate a Z-score. If there are less than two neighbouring residues within 5Ā of the instance then LLDF cannot be calculated. LLDF values of greater than 2 are highlighted.

      B-values: The B-value column is then presented as 4 sub columns providing: the minimum, median, 95th percentile and maximum B-values for all atoms in the instance being analysed.

      This is defined as the number of atoms in the instance which have an occupancy of less than 0.9. In parentheses this value is expressed as a percentage of total observed atoms in the instance.

 

Note that the the LLDF calculation is still under development and the methods and criterion for identifying outlier will be refined following analysis of a large test set of ligands.

 

References

      I. J. Bruno, J. C. Cole, M. Kessler, J. Luo, W. D. S. Motherwell, L. H. Purkis, B. R. Smith, R. Taylor, R. I. Cooper, S. E. Harris, and A. G. Orpen. Retrieval of crystallographically-derived molecular geometry information. J. Chem. Inf. Comput. Sci., 44:2133–2144, 2004.

      A. T. Brunger. Free R value: a novel statistical quantity for assessing the accuracy of crystal structures. Nature, 355 (6359), 472–475, 1992.

      V. B. Chen, W. B. Arendall III, J. J. Headd, D. A. Keedy, R. M. Immormino, G. J. Kapral, L. W. Murray, J. S. Richardson, and D. C. Richardson. MolProbity: all-atom structure validation for macromolecular crystallography. Acta Cryst., D66:12–21, 2010.

      Collaborative Computational Project Number 4. The CCP4 Suite: Programs for Protein Crystallography. Acta Cryst., D50:760–763, 1994.

      R. A. Engh and R. Huber. International Tables for Crystallography, Volume F. Crystallography of Biological Macromolecules., chapter 18.3 Structure quality and target parameters, pages 382– 392. Kluwer Academic Publishers, 2001.

      Z. Feng. Validation-pack. http://sw-tools.pdb.org/

      T. A. Jones, J. Y. Zou, S. W. Cowan and M. Kjeldgaard. Improved methods for building protein models in electron density maps and the location of errors in these models. Acta Cryst. A47, 110-119, 1991.

      G.J. Kleywegt. Validation of protein crystal structures (Topical review). Acta Cryst., D56, 249-265, 2000.

      G. J. Kleywegt, M. R. Harris, J. Zou, T. C. Taylor, A. Wählby, and T. A. Jones. The Uppsala Electron-Density Server. Acta Cryst., D60:2240–2249, 2004.

      J. E. Padilla and T. O. Yeates. A statistic for local intensity differences: robustness to anisotropy and pseudo-centreing and utility for detecting twinning. Acta Cryst., D59:57–64, 2003.

      G. N. Parkinson, J. Vojtechovsky, L. Clowney, A. T. Brünger, and H. M. Berman. New parameters for the refinement of nucleic acid containing structures. Acta Cryst., D52:57–64, 1996.

      R. J. Read. Detecting outliers in non-redundant diffraction data. Acta Cryst., D55:1759–1764, 1999.

      R. J. Read, P. D. Adams, W. B. Arendall III, A. T. Brünger, P. Emsley, R. P. Joosten, G. J. Kleywegt, E. B. Krissinel, T. Lütteke, Z. Otwinowski, A. Perrakis, J. S. Richardson, W. H. Sheffler, J. L. Smith, I. J. Tickle, G. Vriend, and P. H. Zwart. A new generation of crystallographic validation tools for the Protein Data Bank. Structure, 19:1395–1412, 2011.

      H. Yang. DCC. http://sw-tools.pdb.org/

      T. O. Yeates. Simple statistics for intensity data from twinned specimens. Acta Cryst., A44:142–144, 1988.

      wwPDB. The standard geometry compilation used in wwPDB validation protocols, 2012.

      CCD - wwPDB chemical components dictionary, http://www.wwpdb.org/ccd.html