Schemas¶

Kept as-is, or resolved to the best-occupancy/first/ user-defined conformer.

How ties in select_best_occupancy selection are broken.

The altloc code to select when strategy="select_user_defined".

Whether altloc selection diagnostics are recorded.

One AltlocSelectionMappingItem per residue with resolved altlocs.

The canonical chain id the residue belongs to.

The residue's identifier (seq_id:comp_id).

The altloc code that was kept.

Every altloc code that was present before selection.

Why this altloc was selected.

The id assigned to the generated chain.

The original (asymmetric-unit) chain id the copy was generated from.

The id of the symmetry operator applied to produce this copy.

Kept as-is, remapped sequentially, or standardized.

One AssemblyMappingItem per assembly.

The canonicalised assembly id.

The assembly id as originally deposited.

Chains generated by expanding this assembly's symmetry operators, if assembly_rules.strategy == "standardize_biological_assembly".

Kept as reported, standardized to the biological assembly, or only the first assembly.

Which source's "preferred assembly" flag to trust when standardizing.

Whether assembly id mappings are recorded.

The chain id mapping produced by this run.

The residue number mapping produced by this run.

The assembly id mapping produced by this run.

The entity id mapping produced by this run.

The altloc selections made during this run.

The modified residues renamed during this run.

One ChainIdMappingItem per chain.

The canonicalised chain id.

The label_asym_id as originally deposited.

The auth_asym_id as originally deposited.

Kept as-is, remapped sequentially, or replaced by the author chain id.

One EntityMappingItem per canonical entity.

The canonicalised entity id.

The original entity id(s) merged into this canonical id.

Kept as-is, standardized to sequential ids, or merged when sequence-identical.

Whether original entity ids are kept in the mapping when merging.

The chain id canonicalisation policy.

The residue numbering canonicalisation policy.

The assembly id canonicalisation policy.

Kept as-is, excluded, or truncated to the longest contiguous region.

Kept as-is, filtered by keep_waters/keep_ions/ keep_nonpolymer_ligands, or excluded from the canonical structure but annotated.

Whether water molecules are kept.

Whether ions are kept.

Whether other non-polymer ligands are kept.

Kept as-is, flagged, dropped, or imputed.

Whether missing atoms may be imputed (not yet implemented).

Whether missing-atom diagnostics are recorded.

The missing-atoms policy.

The missing-residues (sequence gap) policy.

The incomplete-chain policy.

Kept as-is, flagged, dropped, or imputed.

Whether sequence-gap diagnostics are recorded.

One entry per renamed residue.

The residue's original label_asym_id.

The residue's label_seq_id.

The residue's auth_seq_id.

The modified residue's comp_id (e.g. "MSE").

The standard parent it was renamed to (e.g. "MET").

Kept as-is, or renamed to the parent residue given by _pdbx_struct_mod_residue (MSE always maps to MET).

Which modified residues to map. Empty means every modified residue with a known parent.

One ResidueNumberMappingItem per residue.

The canonical chain id the residue belongs to.

The residue's canonicalised sequence number.

The chain id as originally deposited.

The label_seq_id as originally deposited, if any.

The auth_seq_id as originally deposited.

The insertion code as originally deposited, if any.

Kept as-is, replaced by the author numbering, or renumbered sequentially.

Whether insertion codes are kept when strategy="preserve".

How strictly validation issues are treated ("permissive" is not yet fully implemented).

Whether an unresolved validation error raises instead of just being recorded.

Whether warnings are promoted to errors.

A unique identifier for this policy.

A human-readable name for this policy.

The policy's version string.

A free-text description of the policy.

The chain id / residue numbering / assembly id rules.

The missing-atoms / missing-residues / incomplete-chain rules.

The altloc resolution rules.

The assembly selection rules.

The entity id rules.

The ligand/water/ion filtering rules.

The modified-residue mapping rules.

The validation strictness rules.

Which provenance is recorded for this run.

When this canonicalisation run completed, as an ISO 8601 timestamp.

The id of the policy that ran.

The name of the policy that ran.

The version of the policy that ran.

The transform labels applied (e.g. "chain_id:remap"), one per rule group that deviated from "preserve".

A summary of warning/error counts, if provenance_rules.emit_canonicalisation_report was True.

The verbatim mmCIF categories removed from Structure.raw because they refer to chains or residues by id and this run changed which chains or residues exist.

Whether id/numbering mappings are recorded.

Whether the list of applied transforms is recorded.

Whether the policy id/name/version is recorded.

Whether a warning/error count summary is recorded in provenance.

The assembly this generator instruction belongs to.

The comma-separated operator ids to apply.

The asym units (chains) the operators are applied to.

The operator's id.

The operator type (e.g. "identity operation", "crystal symmetry operation").

The 3x3 rotation matrix.

The 3-element translation vector.

The assembly's id.

Free-text details about the assembly.

The method used to determine the assembly.

A description of the oligomeric state (e.g. "dimeric").

The number of polymer chains in the assembly.

The generator instructions (which operators apply to which chains) that build this assembly.

The symmetry operators referenced by generators.

Whether the depositing author flagged this as the biological assembly.

Whether assembly-detection software (e.g. PISA) flagged this as the biological assembly.

The label_asym_id.

The id of the entity this asym unit belongs to.

The auth_asym_id (author-assigned chain id), if different from id.

"ATOM" for polymer residues, "HETATM" for everything else.

The atom's serial number.

The element symbol.

The atom's name within its residue (e.g. "CA").

The alternate location (altloc) identifier, if the atom has alternate conformers.

The residue/component id (e.g. "ALA").

The asym unit (chain) this atom belongs to.

The entity this atom belongs to.

The residue's sequence position within its entity; None for non-polymer atoms.

The insertion code, if any.

The atom's x coordinate, in angstroms.

The atom's y coordinate, in angstroms.

The atom's z coordinate, in angstroms.

The atom's occupancy (fraction of the site it's present at).

The atom's isotropic (or equivalent) B-factor.

The author-assigned residue sequence number.

The author-assigned residue/component id.

The author-assigned chain id.

The author-assigned atom name.

Which model this atom belongs to, for multi-model entries.

The secondary structure element's id.

The secondary structure type (e.g. "HELX_P", "TURN_P", "STRN").

The chain the element starts in.

The residue sequence position the element starts at.

The chain the element ends in.

The residue sequence position the element ends at.

The author-assigned chain id the element starts in.

The author-assigned residue number the element starts at.

The author-assigned chain id the element ends in.

The author-assigned residue number the element ends at.

The chain of this side of the connection.

The residue/component id of this side of the connection.

The residue sequence position of this side of the connection, if polymeric.

The atom name of this side of the connection.

The author-assigned chain id of this side of the connection.

The author-assigned residue sequence number of this side of the connection.

The insertion code of this side of the connection, if any.

The symmetry operation applied to this side of the connection, if any.

The connection's id.

The connection type: "disulf", "covale", "hydrog", or "metalc".

The first atom in the connection.

The second atom in the connection.

The measured distance between the two atoms, in angstroms.

Free-text details about the connection.

The polymer type (e.g. "polypeptide(L)", "polyribonucleotide"), from _entity_poly.type.

The one-letter sequence as deposited, including any non-standard residue codes in parentheses.

The canonicalized one-letter sequence (non-standard residues mapped to their closest standard equivalent).

The comma-separated auth chain ids this entity spans.

The entity id (label_entity_id).

The entity type: "polymer", "non-polymer", "water", or "branched".

A free-text description of the entity (e.g. protein name).

The entity's molecular formula weight, in daltons.

How the entity was produced (e.g. "man" for genetically manipulated, "nat" for natural, "syn" for synthetic).

Polymer-specific fields, if this entity is a polymer.

The entry's PDB/PDBe identifier (the mmCIF data block name).

The entry's title (_struct.title), if present.

The helix/turn/strand elements from _struct_conf.

The individual beta-sheet strand ranges from _struct_sheet_range.

The beta sheet this strand belongs to.

The strand's id.

The chain the strand starts in.

The residue sequence position the strand starts at.

The chain the strand ends in.

The residue sequence position the strand ends at.

The author-assigned chain id the strand starts in.

The author-assigned residue number the strand starts at.

The author-assigned chain id the strand ends in.

The author-assigned residue number the strand ends at.

The structure's PDB/PDBe entry identifier.

Entry-level identity and title.

Every entity (molecule) in the structure.

Every asym unit (chain), mapping chains to entities.

Every atom in the structure.

Every bond/contact connection between atoms.

Every biological/crystallographic assembly.

The structure's secondary structure elements.

Every other mmCIF category not promoted to a typed field, keyed by category name, preserved verbatim.

A human-readable name for the layer.

A machine-readable type identifying the layer (e.g. "structure_counts").

What the layer describes: an entry, a pair, an interface, etc.

The identifier of the method/algorithm used to compute the layer.

The entry id(s) the layer applies to.

The parameters the layer was computed with.

The layer's actual computed output.

Additional provenance details about how the layer was computed.

The entity's id.

The entity type (e.g. "polymer").

A free-text description of the entity.

The entity's molecular formula weight, in daltons.

How the entity was produced.

The entity's Enzyme Commission number, if applicable.

A description of any mutation(s), if applicable.

A description of the fragment represented, if applicable.

The polymer type, if this entity is a polymer.

The entity's deposited one-letter sequence.

The entity's canonicalized one-letter sequence.

The chain (asym) ids belonging to this entity.

Where this record's data came from.

The structure's entry id.

The entry's title.

The entry's keywords.

The primary citation's title.

The primary citation's DOI.

The primary citation's PubMed id.

Where this record's data came from.

The id of the entity this ligand belongs to, if known.

The ligand's component id (e.g. "ZN").

The ligand's full chemical name.

The ligand's chemical formula.

The ligand's molecular formula weight, in daltons.

The chain (asym) ids this ligand appears in.

Where this record's data came from.

Which source database the value came from.

The mmCIF category the value was read from, if source="mmcif".

The id of the specific record within source_category, if applicable.

The structure's entry id.

Entry-level metadata (title, keywords, citation).

Experimental quality metrics.

Source organism and expression host metadata.

Per-entity metadata.

Per-ligand metadata.

UniProt/SIFTS sequence mappings.

The mmCIF categories that were consulted to build this record.

The experimental method(s) used (e.g. "X-RAY DIFFRACTION").

The structure's resolution, in angstroms, from _refine, or _em_3d_reconstruction for cryo-EM entries.

The working set R-factor.

The free set R-factor.

The merging R-factor as Rsym (_reflns.pdbx_Rsym_value).

The merging R-factor as Rmerge (_reflns.pdbx_Rmerge_I_obs); the same statistic as r_sym under another name.

The number of observed reflections.

The percentage of possible reflections that were observed.

The mean isotropic B-factor.

Where this record's data came from.

The entity this taxonomy applies to.

The source organism's NCBI taxonomy id.

The source organism's scientific name.

The source organism's common name.

The expression host's NCBI taxonomy id, if genetically manipulated.

The expression host's scientific name.

The expression host's common name.

The expression system or plasmid used.

Where this record's data came from.

The entity this mapping applies to, if known.

The chain this mapping applies to, if known.

The UniProt accession.

The reference database name.

The reference database entry code.

The mapping's start position in the structure's sequence numbering.

The mapping's end position in the structure's sequence numbering.

The mapping's start position in the UniProt sequence.

The mapping's end position in the UniProt sequence.

The sequence identity of the mapped segment, if reported (SIFTS mappings only).

Where this record's data came from.

A human-readable name for the layer.

A machine-readable type identifying the layer (e.g. "structure_counts").

What the layer describes: an entry, a pair, an interface, etc.

The identifier of the method/algorithm used to compute the layer.

The entry id(s) the layer applies to.

The parameters the layer was computed with.

The layer's actual computed output.

Additional provenance details about how the layer was computed.

Whether a batch fetch continues past individual failures instead of raising.

Whether an existing cached file is reused instead of refetching.

Whether a gzip-compressed response is decompressed before writing.

How old a cached file may be before it's considered stale; None means never stale.

What to do with a stale cached file: use it anyway, warn and refetch, or raise.

Which provider the file was fetched from ("pdbe" or "pdb").

The exact URL the file was fetched from.

When the file was retrieved, as an ISO 8601 timestamp.

Whether the file came from the local cache rather than a fresh download.

The entry's own most recent revision date (ISO "YYYY-MM-DD"), when the provider exposes one. Lets a caller detect drift on a later re-fetch of the same entry_id without Pandora computing a content checksum. None when the provider doesn't expose revision info or it couldn't be determined.

The policy's unique identifier.

The policy's human-readable name.

The policy's version string.

A short machine-readable code identifying the issue (e.g. "RESIDUE_NUMBER_COLLISION").

"warning" or "error".

A human-readable description of the issue.

The entry this diagnostic applies to, if known.

The field/category this diagnostic relates to, if applicable.

Additional structured details about the issue.

Every warning-level diagnostic collected.

Every error-level diagnostic collected.

The structure's entry id.

The chain's label_asym_id.

The number of positions in the chain entity's _entity_poly_seq.

Missing positions before the first present residue.

Missing positions after the last present residue.

Missing positions between the first and last present residue.

The structure's entry id.

The chain's label_asym_id.

The chain's auth_asym_id, if different.

The id of the entity this chain belongs to.

The chain's polymer sequence, if any.

The number of distinct residues in the chain.

The number of atoms in the chain.

Whether non-polymer ligands are kept in curated structures.

Whether water molecules are kept in curated structures.

Whether ions are kept in curated structures.

When this curation run completed, as an ISO 8601 timestamp.

The id of the policy that ran.

The name of the policy that ran.

The version of the policy that ran.

The verbatim mmCIF categories removed from Structure.raw because they refer to chains or residues by id and curation removed some chains or ligands.

A unique identifier for this policy.

A human-readable name for this policy.

The policy's version string.

A free-text description of the policy.

The resolution/method/chain-length rules.

The organism taxon inclusion/exclusion rules.

The ligand/water/ion content rules.

When this deduplication run completed, as an ISO 8601 timestamp.

Whether deduplication was actually applied.

How many structures were removed as duplicates.

Whether structures sharing the same entry_id are removed.

The excluded structure's entry id.

The excluded chain's label_asym_id, or None when the whole entry was excluded.

A machine-readable reason the structure was excluded.

A human-readable description of the exclusion.

The structure's entry id.

The first chain in the interface.

The second chain in the interface.

The contact distance cutoff used, in angstroms.

Which atoms counted towards contacts ("all", "heavy", "backbone" or "ca"). Records written before this field existed load as "all", which is what they used.

Contacting residue ids on chain_id_1's side.

Contacting residue ids on chain_id_2's side.

The total number of contacting residues on both sides.

Contacting (chain_id_1 residue, chain_id_2 residue) id pairs, sorted.

If non-empty, only these NCBI taxon ids are allowed.

NCBI taxon ids that are always excluded.

The maximum resolution (in angstroms) allowed; structures coarser than this are excluded.

Whether a structure with no resolution is excluded or included when max_resolution is set.

The minimum residue count a structure's longest chain must reach, checked after chain rules remove chains.

The minimum number of polymer chains (asym units) a structure must have, checked after chain rules remove chains.

If non-empty, only these experimental methods are allowed.

Experimental methods that are always excluded.

Resolution limits per experimental method, overriding max_resolution for entries with that method. With several matching methods, the strictest wins.

The methods the R-factor rules apply to.

The maximum Rfree.

The maximum |Rfree - Rwork|.

The maximum Rsym, or Rmerge when Rsym is missing.

Whether an entry missing a value an active R-factor rule needs is excluded or included.

Whether an entry with any non-standard polymer residue is excluded.

Non-standard residues that don't trigger exclude_nonstandard_residues.

The maximum number of atoms an entry may have.

What counts as a missing residue for the missing-fraction rules (see chain_completeness).

Chains whose missing N- plus C-terminal residues exceed this fraction of SEQRES are removed.

Chains whose missing middle residues exceed this fraction of SEQRES minus the missing tails are removed.

Chains with more observed residues than this are removed.

The structure's entry id.

The residue's label_asym_id.

The residue's label_seq_id, if polymeric.

The residue's auth_seq_id.

The residue's component id (e.g. "ALA").

Every atom belonging to this residue.

The annotation layer's human-readable name.

The annotation layer's machine-readable type (e.g. "ligand_contacts").

The identifier of the method/algorithm used to compute the layer.

The entry id(s) the layer applies to.

The parameters the layer was computed with.

Additional provenance details about how the layer was computed.

A unique identifier for the dataset.

A human-readable name for the dataset.

The dataset's version string.

The Pandora version that produced this manifest.

When this manifest was assembled, as an ISO 8601 timestamp.

The curation policy applied, if any.

The canonicalisation policy applied, if any.

Every structure excluded during curation or deduplication.

How deduplication was applied, if any.

How similarity clustering was applied, if any.

How the train/val/test split was computed, if any.

The resulting split assignment, keyed by split name.

The ProvenanceBundle for every retained structure.

The structure's entry id.

The Pandora version that produced this bundle.

When this bundle was assembled, as an ISO 8601 timestamp.

How the structure's raw file was fetched, if known.

How the structure was canonicalised, if known.

Which sources the structure's metadata came from.

Every annotation layer computed for this structure.

When this clustering run completed, as an ISO 8601 timestamp.

The filter that turned hits into edges, or None for hand-built edges (cluster_edges).

The search whose hit file was clustered, or None for hand-built edges.

How many edges were read (duplicates included).

How many of those named an id outside item_ids and were ignored.

How many clusters resulted.

The lexicographically smaller item id.

The lexicographically larger item id.

The row's score, as selected by the filter.

The row's fraction of identical aligned residues.

The row's coverage, as selected by coverage_of.

Interface-restricted coverage, or None if not computed for this pair.

Minimum score: TM-score for Foldseek (see tm_normalisation), identity for MMseqs2.

Minimum fraction of identical aligned residues.

Minimum alignment coverage (see coverage_of).

Which coverage is tested: "both" (the smaller of query and target), "query", "target", or "either" (the larger).

Minimum fraction of each side's interface residues inside the alignment (Foldseek only; needs interface_residues).

Foldseek TM-score used as the score: "alignment" (alntmscore), "query" (qtmscore), "target" (ttmscore), or "max" (the larger of the two).

When this partition run completed, as an ISO 8601 timestamp.

The target fraction of items assigned to train.

The target fraction of items assigned to val.

The target fraction of items assigned to test.

Whether each cluster was kept whole in one split (leakage-safe) or divided proportionally.

The resulting number of items per split.

Every item id in this cluster.

The number of items in this cluster.

The similarity engine used (e.g. "MMseqs2", "Foldseek").

The engine's version string, if determined.

The parameters the engine was run with.

The lexicographically smaller of the two item ids.

The lexicographically larger of the two item ids.

What kind of similarity this measures.

The relationship's primary similarity score.

The alignment coverage between the two items, if reported.

The alignment coverage restricted to each item's interface residues, if computed (see compute_structure_similarity(interface_residues=...)).

The sequence/structural identity between the two items, if reported.

Which engine/parameters produced this relationship.

When and for which dataset this relationship was computed.

When this relationship was computed, as an ISO 8601 timestamp.

The dataset this relationship was computed for, if known.

Which tool produced the hits.

The tool's version string, if determined.

Absolute path to the hit TSV, so the record works from any working directory.

The TSV's columns, in order.

The search settings (binary, sensitivity, max_seqs, extra options, ...), as passed to the search function, so the search can be re-run.

"computed" if Pandora ran the search, "precomputed" if an existing file was wrapped with load_similarity_search().

When the search ran, as an ISO 8601 timestamp.