PDB-IHM

1. Understanding the PDB-IHM Validation Report

This validation report was created based on the guidelines and recommendations from IHM TaskForce (Berman et al. 2019). The current version of the PDB-IHM validation report consists of four categories as follows:

1.1 Model composition: This section outlines model details and includes information on ensembles deposited, chains and residues of domains, model representation, software, protocol, and methods used. All deposited structures have this section.

1.2. Data quality assessment: Data quality assessments are available for Small Angle Scattering datasets (SAS) and Chemical Crosslinking Mass Spectrometry (crosslinking-MS) data and is based on the guidelines published by the wwPDB SAS validation task force (Trewhella et al. 2017) and crosslinking-MS community (Leitner et al. 2020).

1.3. Model quality assessment: Model quality for models at atomic resolution is assessed using MolProbity (Williams et al. 2018), consistent with PDB. Model quality for coarse-grained or multi-resolution structures are assessed by computing excluded volume satisfaction based on reported distances and sizes of beads in the structures. Model precision, defined as the variability among the models that satisfy the input data and calculated as the density-weighted root mean-square fluctuation (RMSF) from the bead/atom center of density, is annotated and visualized using PrISM (Ullanat et al. 2022). Only coarse-grained beads (or CA atoms for atomic models) of deposited models are used for precision assessment and visualization.

1.4. Fit to data used to build the model: Fit to data used to build the model is available for SAS and crosslinking-MS datasets. This section was developed in collaboration with the SAS and crosslinking-MS communities and SASBDB (Valentini et al. 2015) and PRIDE (Perez-Riverol et al. 2025) resources. For details on the metrics, guidelines, and recommendations used, refer to the community guidelines (Trewhella et al. 2017, Leitner et al. 2020). All experimental datasets used to build the model are listed, however, validation criteria for other types of experimental data are currently under development.

A fifth category, fit to data used to validate the model, is under development.

2. Overview

2.1 Overall Quality Assessment: This is a set of plots that represent a snapshot view of the validation results. There are four tabs, one for each validation criterion: (i) model quality, (ii) data quality, (iii) fit to data used for modeling, and (iv) fit to data used for validation.

• 2.1.1. Model quality: For atomic structures, MolProbity is used for evaluation. We evaluate bond outliers, side chain outliers, clash score, rotamer satisfaction, and Ramachandran dihedral satisfaction (Williams et al. 2018) . Details on MolProbity evaluation and tables can be found here. For coarse-grained structures of beads, we evaluate excluded volume satisfaction. An excluded volume violation or overlap between two beads occurs if the distance between the two beads is less than the sum of their radii (S. J. Kim et al. 2018). Excluded volume satisfaction is the percentage of pair distances in a structure that are not violated (higher values are better).

• 2.1.2. Data quality: Data quality assessments are available for SAS and crosslinking-MS datasets. The current plot displays radius of gyration (R_g) for each SAS dataset used to build the model. R_g is obtained from both a P(r) analysis (see more here), and a Guinier analysis (see more here). For the crosslinking-MS datasets we assess consistency between experimental data and data used for modeling (see more here).

• 2.1.3. Fit to data used for modeling: Fit to data used for modeling assessments are available for SAS and crosslinking-MS datasets. The plot displays Χ² Goodness of Fit Assessment for SAS-model fits (see more here) and percentage of satisfied crosslinking-MS restraints (see more here).

• 2.1.4. Fit to data used for validation: Fit to data used for validation is currently under development.

3. Model Details

3.1. Ensemble Information: Number of ensembles deposited, where each ensemble consists of two or more structures.

3.2. Summary: Summary of the structure, including number of models deposited, datasets used to build the models and information on model representation.

3.3. Entry Composition: Number of chains present in the integrative structure.

3.4. Datasets Used: Number and type of experimental datasets used to build the model.

3.5. Representation: Number and details on rigid and non-rigid elements of the structure.

3.6. Methods and Software: Methods, protocols, and softwares used to build the integrative structure.

4. Data Quality

4.1. SAS: Scattering Profiles: Data from solutions of biological macromolecules are presented as both log I(q) vs q and log I(q) vs log (q) based on SAS validation task force (SASvtf) recommendations (Trewhella et al. 2017). I(q) is the intensity (in arbitrary units) and q is the modulus of the scattering vector.

4.2. SAS: Experimental Estimates: Molecular weight (MW) and volume data are displayed. True MW can be compared to Porod estimate from scattering profiles, estimated volume can be compared to Porod volume obtained from scattering profiles (Trewhella et al. 2017).

4.3. SAS: Flexibility Analysis: Flexibility of chains are assessed by inferring Porod-Debye and Kratky plots. In a Porod-Debye plot, a clear plateau is observed for globular (partial or fully folded) domains, whereas fully unfolded domains are devoid of any discernible plateau. For details, refer to Figure 5 in Rambo and Tainer, 2011 (Rambo and Tainer 2011). In a Kratky plot, a parabolic shape is observed for globular (partial or fully folded) domains and a hyperbolic shape is observed for fully unfolded domains.

4.4. SAS: P(r) Analysis: P(r) represents the distribution of distances between all pairs of atoms within the particle weighted by the respective electron densities (Moore 1980) . P(r) is the Fourier transform of I(s) (and vice versa). R_g can be estimated from integrating the P(r) function. Agreement between the P(r) and Guinier-determined R_g (table below) is a good measure of the self-consistency of the SAS profile. R_g is a measure for the overall size of a macromolecule; e.g. a protein with a smaller R_g is more compact than a protein with a larger R_g, provided both have the same molecular weight (MW). The point where P(r) is decaying to zero is called D_max and represents the maximum size of the particle. The value of P(r) should be zero beyond r=D_max.

4.5. SAS: Guinier Analysis: Agreement between the P(r) and Guinier-determined R_g (table below) is a good measure of the self-consistency of the SAS profile. The linearity of the Guinier plot is a sensitive indicator of the quality of the experimental SAS data; a linear Guinier plot is a necessary but not sufficient demonstration that a solution contains monodisperse particles of the same size. Deviations from linearity usually point to strong interference effects, polydispersity of the samples or improper background subtraction (Feigin and Svergun 1987). Residual value plot and coefficient of determination (R²) are measures to assess linear fit to the data. A perfect fit has an R² value of 1. Residual values should be equally and randomly spaced around the horizontal axis.

4.6. Crosslinking-MS: At present, data validation is only available for crosslinking-MS data deposited as a fully compliant dataset in PRIDE database. Correspondence between crosslinking-MS and entry entities is established using pyHMMER. Only residue pairs that passed the reported threshold are used for the analysis. The values in the report have to be interpreted in the context of the experiment (i.e. only a minor fraction of in-situ or in-vivo dataset can be used for modeling).

5. Model Quality Assessment

Excluded volume assessments are performed for coarse-grained structures and MolProbity analysis is performed for atomic structures.

5.1a. Excluded Volume Analysis: Excluded volume violation is defined as percentage of overlaps between coarse-grained beads in a structure. This percentage is obtained by dividing the number of overlaps/violations by the total number of pair distances in a structure. An overlap or violation between two beads occurs if the distance between the two beads is less than the sum of their radii (S. J. Kim et al. 2018).

5.1b. MolProbity Analysis: MolProbity analysis for atomic structures reported is consistent with PDB standards for X-ray structures (Williams et al. 2018). Summarized information is available in both the HTML and PDF reports. Detailed information is available for download as csv files, both from the HTML and the PDF reports. Please refer to the PDB user guide for details.

5.2. PrISM Precision Analysis: Regions of low and high precision, defined as the variability among the models that satisfy the input data and calculated as the density-weighted root mean-square fluctuation (RMSF) from the bead/atom center of density, are annotated and visualized using PrISM (Ullanat et al. 2022). The per-bead or per-residue precision is computed from the deposited ensemble of superposed integrative models. High- and low-precision regions are then determined by clustering beads of similar precision based on their proximity in the structure. Only coarse-grained beads (or CA atoms for atomic models) of deposited models are used for precision assessment and visualization, and three projections for each representative model are generated.

6. Fit to Data Used for Modeling Assessment

Recommendations from SAS validation task force (SASvtf) for model fit assessment include:

6.1. SAS: Χ² Goodness of Fit Assessment: Model fits displayed in this section are obtained from SASBDB. χ² values are a measure of fit of the model to data. A perfect fit has a χ² value of 1.0. (Trewhella et al. 2013, Schneidman-Duhovny, Kim, and Sali 2012, and Rambo and Tainer 2013).

6.2. SAS: Cormap Analysis: ATSAS DATCMP (Manalastas-Cantos et al. 2021) was used for hypothesis testing, using the null hypothesis that all data sets (i.e. the fit and the data collected) are similar. The reported p-value is a measure of evidence against the null hypothesis; the smaller the value, the stronger the evidence that the null hypothesis should be rejected.

6.3. Crosslinking-MS: Restraint types: Summary table of crosslinking-MS restraint types reported in the entry and distogram plots. Restraints assigned "by-residue" are interpreted as between CA atoms. Restraints between coarse-grained beads are indicated as "coarse-grained". Restraint group represents a set of crosslinking restraints applied collectively in the modeling. Restraints with identical thresholds are grouped into one plot. Only the best distance per restraint per model group/ensemble is plotted. Inter- and intramolecular (including self-links) restraints are also grouped into one plot. Distance for a restraint between coarse-grained beads is calculated as a minimal distance between shells; if beads intersect, the distance will be reported as 0.0. A bead with the highest available resolution for a given residue is used for the assessment.

6.4. Crosslinking-MS: Satisfaction rates: Satisfaction of restraints is calculated on a restraint group (a set of crosslinking restraints applied collectively in the modeling) level. Satisfaction of a restraint group depends on satisfaction of individual restraints in the group and the conditionality (all/any). A restraint group is considered satisfied, if the condition was met in at least one model of the model group/ensemble. Only deposited models are used for validation right now.

7. Fit to Data Used for Validation Assessment

This includes assessing model fit to data that was not used explicitly or implicitly in modeling. This section is currently under development.

8. Understanding the Summary Table

8.1. Entry composition: List of unique molecules that are present in the entry.

8.2. Datasets used for modeling: List of input experimental datasets used for modeling.

8.3. Representation: Representation of modeled structure.

• 8.3.1. Scale: Types and sizes of geometric objects comprising structural models.

• 8.3.2. Atomic structural coverage: Percentage of modeled structure or residues for which atomic structures are available. These structures can include X-ray, NMR, EM, and other comparative models.

• 8.3.3. Rigid segment: A rigid segment consists of multiple coarse-grained (CG) beads or atomic residues. In a rigid segment, the beads (or residues) have their relative distances constrained during conformational sampling.

• 8.3.4. Flexible segment: Flexible segments consist of strings of beads that are restrained by the sequence connectivity.

• 8.3.5. Interface units: An automatic definition based on identified interface for each model. Applicable to models built with HADDOCK.

• 8.3.6. Resolution: An automatic definition based on identified interface for each model. Applicable to models built with HADDOCK.

8.4. Restraints: A set of restraints used to compute modeled structure.

• 8.4.1. Physical restraints: A list of restraints derived from physical principles to compute modeled structure.

• 8.4.2. Experimental information: A list of restraints derived from experimental datasets to compute modeled structure.

8.5. Validation: Assessment of models based on validation criteria set by IHM task force (Sali et al. 2015 and Berman et al. 2019).

• 8.5.1. Sampling validation: Validation metrics used to assess sampling convergence for stochastic sampling. Sampling precision is defined as the largest allowed Root-mean-square deviation (RMSD) between the cluster centroid and a model within any cluster in the finest clustering for which each sample contributes structures proportionally to its size (considering both the significance and magnitude of the difference) and for which a sufficient proportion of all structures occur in sufficiently large clusters (Viswanath et al. 2017).

• 8.5.2. Clustering algorithm: Clustering algorithm used to analyze resulting solution.

• 8.5.3. Clustering feature: Feature or reaction co-ordinate used to cluster solution.

• 8.5.4. Number of ensembles: Number of solutions or ensembles of modeled structure.

• 8.4.5. Number of models in ensemble(s): Number of structures in the solution ensemble(s).

• 8.5.6. Model precision: Measurement of variation among the models in the ensemble upon a global least-squares superposition.

• 8.5.7. Data quality:Assessment of data on which modeled structures are based. See section 4 for more details.

• 8.5.8. Model quality:Assessment of modeled structures based on physical principles.See section 5 for more details.

• 8.5.9. Assessment of atomic segments:Assessment of atomic segments in the integrative structure. See section 5 for more details.

• 8.5.10. Excluded volume satisfaction:Assessment of excluded volume satisfaction of coarse-grained beads in the modeled structure. Excluded volume between two beads not connected in sequence are satisfied if the distance between them is greater than that of the sum of their radii. See section 5 for more details.

• 8.5.11. Fit to data used for modeling:Assessment of modeled structure based on data used for modeling. See section 6 for more details.

• 8.5.12. Fit to data used for validation:Assessment of modeled structure based on data not used for modeling. See section 7 for more details.

8.6. Methodology and software: List of methods on which modeled structures are based and software used to obtain structures.

• 8.6.1. Name: Name(s) of the modeling step(s).

• 8.6.2. Method: Name(s) of method(s) used to generate modeled structures.

• 8.6.3. Method description: Details of method(s) used to generate modeled structures.

• 8.6.4. Number of computed models: Number of models computed at each modeling step.

• 8.6.5. Software details: Software used to compute modeled structure, also includes scripts used to generate and analyze models.

9. References for Validation Report