{ "dataset": "SMARTFlexDB apo-holo pair table", "schema_version": 12, "data_file": "smartflexdb_pairs.tsv", "generated_utc": "2026-09-22T18:38:28Z", "source": { "file": "data_20260919b/showcase.json", "sha256": "dff04c210c0c879295d3df166f4bb919f03eb6ada37c342bde27a68cd424c4e7" }, "tool_manifest": "/api/versions", "tool_manifest_generated_utc": "2026-09-21T06:08:02+00:00", "input_list_sha256": "ffd57b81c5a374474449aa114098f3bac01b0a101a9fb387fdf53f798a35ac6f", "rows": 1187, "rows_by_apo_qualified": { "qualified": 1187, "unqualified": 0, "unjudged": 0 }, "changes": { "schema_version_from": 11, "schema_version_to": 12, "date": "2026-09-22", "items": [ "one row per POCKET-LEVEL COMPARISON - one binding pocket against one apo structure it is compared with (was: one row per apo-holo entry and host RNA strand): 1,187 rows instead of 794, every served binding pocket has its own row, and `pocket_rmsd`, `pocket_size`, `hbonds`, `mw`, `logp`, `lig_resseq` and the two motif columns are that binding pocket's own", "`ligand` is this binding pocket's own ligand (was: every ligand on the row's host RNA strand, joined with '; '), so the column holds one CCD code", "`id` is unchanged (`__apo_`) and is now the row key on its own; group on (holo_pdb, apo_pdb) for the 734 apo-holo entries", "sidecar: `row_unit.collapse_is_lossy` and `served_records.binding_pockets_without_a_row` are empty by construction; `row_unit.representative` says there is none", "REMOVED at schema_version 11 (2026-09-19): the `apo_role` column (33 -> 32 columns)", "REMOVED at schema_version 11 (2026-09-21): the rows of apo structures the sequence check refused" ], "items_schema_version_11": [ "column `apo_role` removed (33 -> 32 columns); `holo_rna_chain` moves from column 33 to column 32, every other column keeps its name, meaning and order", "`id` is the entry id `__apo_` on every row, the binding pocket's own apo structure included (was the bare binding pocket id, or `__alt_`); /api/pair/{id}, /api/bundle/{id} and /api/thumb/{id} accept it, and the old forms still work", "sidecar: `rows_by_apo_role` removed; `rows_by_apo_qualified` is one count per verdict; `served_records.ids_without_a_default_row` replaced by `binding_pockets_without_a_row`; `row_unit.unqualified_apo_states` renamed `unqualified_apo_structures`; `row_unit.alternatives_deduped_by` renamed `apo_deduped_by`; `row_unit.collapse_is_lossy` reports one census, `combinations`; this `changes` block added", "rows in ascending order of `id`: a binding pocket's rows together, by apo PDB ID (was: grouped, each group opening with one designated apo structure)", "sidecar: `apo_copies_emitted` and `apo_copies_present` removed; the `apo_chain` column note states which chain is used when the apo structure holds the same RNA chain several times", "rows whose apo structure failed the sequence check removed (2026-09-21; they were kept and marked `apo_qualified = unqualified`): every row is an entry the site shows, and `apo_qualified` now reads `qualified` or `unjudged`" ] }, "row_unit": { "unit": "one row per POCKET-LEVEL COMPARISON: one binding pocket against one apo structure", "key": [ "id" ], "key_columns": [ "id" ], "key_note": "`id` is `__apo_` and is unique: the binding pocket before the infix, the apo structure after it. Group on (`holo_pdb`, `apo_pdb`) for the apo-holo ENTRIES of the release, and count distinct binding-pocket ids for the BINDING POCKETS. `holo_chain` (published again as `holo_rna_chain`, equal on every row; measured equal on 434 of 434 served records at emit time) is the host strand the geometry belongs to", "definition": "backend/export_dataset.py::plan_rows (one row per served binding pocket and offered apo structure; the same population as /api/pairs)", "representative": "none: every served binding pocket has its own row, and every cell is that binding pocket's own", "apo_deduped_by": "apo PDB id: a binding pocket has one row per distinct apo structure it is compared with", "changed": "2026-08-15 (schema_version 7 -> 8): rows used to be one per served POCKET RECORD, which made this file disagree with the Database table it is downloaded from; 2026-09-14 (schema_version 9 -> 10): re-keyed from (holo_pdb, ligand, apo_pdb) to (holo_pdb, apo_pdb, holo_rna_chain); 2026-09-17: vocabulary only - the (holo_pdb, apo_pdb) pair, formerly called a comparison, is now called an ENTRY, and the row set is unchanged; 2026-09-19 (schema_version 10 -> 11): the row set is unchanged, see `changes`; 2026-09-21: the rows whose apo structure failed the sequence check were removed, see `changes`", "unqualified_apo_structures": { "ruling": "an apo structure whose sequence-qualification verdict is FALSE is not offered on the Database page or a pair page and has no row in this download (owner ruling 2026-09-21; from 2026-09-13 until then such rows were kept and flagged 'unqualified' in the apo_qualified column)", "rule": "pipeline/seq_qualification.py judges it; pipeline/serve_filter.py::seq_unqualified is the serve gate, and it fires on `is False` and nothing else - an UNJUDGED apo structure is not a refused one", "read_it_as": "every row is an entry the site shows; no filtering is needed to reproduce the site's entries", "entries_not_in_this_file": 23 }, "collapse_is_lossy": { "note": "no binding pocket shares another's row: the unit is the pocket-level comparison, so every served binding pocket is published with its own values", "combinations": { "dropped": 0, "differ_from_the_row_that_survives": 0, "columns_that_move": {} } } }, "served_records": { "served": 434, "built": 761, "rule": "pipeline/serve_filter.py::serve_hide_reasons", "without_a_row": 0, "note": "434 served binding pockets, 1187 rows: one row per binding pocket and apo structure it is compared with, so every served binding pocket has a row of its own. The rows cover the 734 apo-holo entries the site shows, over 266 holo structures; the 23 (holo, apo) pairings the sequence check refused have no row", "binding_pockets_without_a_row": [] }, "row_order": "rows in ascending order of id (__apo_): a binding pocket's rows together, by apo PDB ID", "binding_site": { "cutoff_angstrom": 8.0, "definition": "any ligand heavy-atom contact <= 8.0 A, across EVERY contacting RNA chain" }, "license": "CC-BY-4.0 (derived data); coordinates from the RCSB PDB", "read_with": { "pandas": "read_csv(path, sep=\"\\t\", comment=\"#\")", "R": "read.table(path, sep=\"\\t\", header=TRUE)" }, "filter_state": null, "columns": [ { "name": "id", "unit": "key", "side": "entry", "chain_scope": null, "note": "the entry id __apo_ - this row's binding pocket and its apo structure, and the key of the row (unique). /api/pair/{id} and /api/bundle/{id} accept it; the binding pocket in the id is the one this row measures: since 2026-09-22 every served binding pocket has its own row against every apo structure it is compared with, so no row speaks for another binding pocket (1187 rows in this release)" }, { "name": "holo_pdb", "unit": "PDBID", "side": "holo", "chain_scope": "structure" }, { "name": "holo_chain", "unit": "auth", "side": "holo", "chain_scope": "host", "note": "RNA HOST chain (= holo_rna_chain; at-rest R6)" }, { "name": "lig_auth_chain", "unit": "auth", "side": "holo", "chain_scope": "ligand instance", "note": "the chain the LIGAND sits on; identical to holo_chain on all 434 served pockets in this release - since the 2026-07-31 host-chain repoint the builder sets the RNA host chain from the ligand's auth chain, so a join on either column attributes the same way" }, { "name": "ligand", "unit": "CCD", "side": "holo", "chain_scope": "ligand instance", "note": "the CCD code of THIS binding pocket's ligand - one code per row since schema_version 12 (2026-09-22), because a row is one binding pocket. The cell does NOT depend on which apo the row names. The Database page's own CSV has a column of the same name whose row unit is the ENTRY - one (holo_pdb, apo_pdb) - and whose cell lists every ligand of the HOLO STRUCTURE, joined with '; ', so it is a superset of this one (0 of this release's 1187 rows carry more than one code; the unit is one binding pocket, so the column holds a single ligand)" }, { "name": "lig_resseq", "unit": "auth#", "side": "holo", "chain_scope": "ligand instance" }, { "name": "apo_pdb", "unit": "PDBID", "side": "apo", "chain_scope": "structure" }, { "name": "apo_chain", "unit": "auth", "side": "apo", "chain_scope": "host", "note": "the apo chain paired to holo_chain by the superposition; where the apo structure holds the same RNA chain several times, the one with the fewest missing residues" }, { "name": "rna_type", "unit": "class", "side": "holo", "chain_scope": "host entity", "note": "NAKB curated functional class; \"Unannotated\" where NAKB annotates the host entity with none" }, { "name": "method", "unit": "enum", "side": "holo", "chain_scope": "structure" }, { "name": "resolution", "unit": "A", "side": "holo", "chain_scope": "structure" }, { "name": "identity", "unit": "%", "side": "pair", "chain_scope": "ligand auth chain", "note": "BLAST pident - ALIGNED REGION ONLY" }, { "name": "global_identity", "unit": "%", "side": "pair", "chain_scope": "ligand auth chain", "note": "Needleman-Wunsch over DECLARED SEQRES; every mismatch AND gap column counts" }, { "name": "holo_len", "unit": "nt", "side": "holo", "chain_scope": "host", "note": "RNA-polymer length of the host chain only (gemmi CCD is_nucleic_acid OR the entry's own _entity_poly declaration) - NOT the modeled/SEQRES count the pair page's 3D-incompleteness warning uses" }, { "name": "apo_len", "unit": "nt", "side": "apo", "chain_scope": "host", "note": "RNA-polymer length of the host chain only (gemmi CCD is_nucleic_acid OR the entry's own _entity_poly declaration) - NOT the modeled/SEQRES count the pair page's 3D-incompleteness warning uses" }, { "name": "num_chains", "unit": "count", "side": "holo", "chain_scope": "pocket", "note": "RNA chains with a contact <= 8 A" }, { "name": "has_protein", "unit": "bool", "side": "holo", "chain_scope": "structure", "note": "residue composition (gemmi CCD), amino-acid majority" }, { "name": "has_dna", "unit": "bool", "side": "holo", "chain_scope": "structure", "note": "carries a MAJORITY-DEOXY chain (gemmi CCD residue composition, strict > 50%) - a minority of 2'-deoxy residues in an RNA oligo does NOT count; the depositor's _entity_poly.type is not used" }, { "name": "rmsd", "unit": "A", "side": "pair", "chain_scope": "all pocket chains", "note": "POOLED C3' RMSD over the chains in the fit - every <= 8 A pocket chain with a distinct injective apo counterpart - in the one published superposition frame; reduces to the host chain alone wherever the fit had a single chain to solve on, which showcase.json marks fit_mode=host_only (single-chain pockets, pockets whose strands would share one apo chain, and a pocket whose second chain has no injective apo counterpart)" }, { "name": "host_rmsd", "unit": "A", "side": "pair", "chain_scope": "host", "note": "the ligand host chain (lig_auth_chain) measured IN THE PUBLISHED all-chain frame; == rmsd for a single-chain pocket, larger than host_rmsd_local when the pocket's chains have moved relative to each other" }, { "name": "host_rmsd_local", "unit": "A", "side": "pair", "chain_scope": "host", "note": "the ligand host chain's OWN best fit (frame-independent) - the successor of the pre-2026-08-05 `host_chain_rmsd` column, which was that same host-only fit. host_rmsd - host_rmsd_local is the inter-chain hinge signal" }, { "name": "tm_score", "unit": "0-1", "side": "pair", "chain_scope": "host", "note": "US-align" }, { "name": "pocket_rmsd", "unit": "A", "side": "pair", "chain_scope": "all pocket chains", "note": "RMS C3' shift over the <= 8 A binding-site residues" }, { "name": "pocket_size", "unit": "count", "side": "holo", "chain_scope": "all contacting chains", "note": "interactions3d rows <= 8 A" }, { "name": "hbonds", "unit": "count", "side": "pair", "chain_scope": "all contacting chains", "note": "H-BONDED BINDING-SITE RESIDUES (interactions3d rows typed hbond) - NOT the atom-level DSSR donor/acceptor pairs the pair page's Hydrogen-bonds tab lists; that count is len(dssr_hbonds) on /api/pair and differs on 307 of the 404 served records that carry a non-empty dssr_hbonds (larger on 307, smaller on 0); the other 30 carry none, so the tab falls back to these very rows" }, { "name": "mw", "unit": "Da", "side": "ligand", "chain_scope": null }, { "name": "logp", "unit": "--", "side": "ligand", "chain_scope": null }, { "name": "rg_delta", "unit": "A", "side": "pair", "chain_scope": "all pocket chains", "note": "rg_holo - rg_apo, each the radius of gyration of the C3' atoms of the apo<->holo ALIGNED CORE of every chain in the superposition fit set, POOLED into one point cloud per state - the same residues `rmsd` is measured over, so the three descriptors describe the same object. On a multi-chain pocket a near-zero value therefore means the whole binding unit kept its SIZE, not that each strand did: an inter-chain rearrangement now enters rg_delta as it already enters rmsd" }, { "name": "pocket_motifs", "unit": "list", "side": "holo", "chain_scope": "host", "note": "x3dna-dssr motif types whose residues overlap the <= 8 A binding site, \"; \"-joined - the COARSE level, in which every N-way junction is collapsed to the single value \"junction\". THIS is the value the Database facet, the Statistics \"Pocket motifs\" bars and every ?motif= link filter on, so it is the column that reproduces a filtered selection; pocket_motifs_fine carries the subtype" }, { "name": "pocket_motifs_fine", "unit": "list", "side": "holo", "chain_scope": "host", "note": "the SAME computation with DSSR's raw type kept, so a junction reads \"4-way junction\"; identical to pocket_motifs on a pocket with no junction. A LABEL SET, NOT A PARTITION of the coarse value - measured at emit time over the 434 served binding pockets: the 219 that carry a junction carry 324 subtypes between them, so summing per-subtype counts overcounts the junction population by 105 - count distinct rows. This file's rows are binding-pocket comparisons, so a binding pocket is counted once per apo structure it is compared with: 627 of its 1187 rows carry a junction, 1030 subtypes between them" }, { "name": "apo_qualified", "unit": "enum", "side": "pair", "chain_scope": "all pocket chains", "note": "the SEQUENCE-QUALIFICATION verdict of THIS ROW's apo structure (pipeline/seq_qualification.py): qualified | unqualified | unjudged - every binding-site chain has a paired apo chain whose 8 A site positions all match, and the two author-declared assemblies carry the same RNA strand count (that half is a WHOLE-ASSEMBLY check, wider than this column's chain scope). UNJUDGED IS NOT UNQUALIFIED: only an explicit refusal is one, and a row this release did not judge must not be read as failing - the serve rule (pipeline/serve_filter.py::seq_unqualified) hides on `is False` and on nothing else. An apo structure the check refused is not offered on the Database page or a pair page and has no row here either (owner ruling 2026-09-21), so every row is an entry the site shows; measured at emit time - of this release's 1187 rows 1187 are qualified, 0 are unqualified and 0 are unjudged" }, { "name": "holo_rna_chain", "unit": "auth", "side": "holo", "chain_scope": "host", "note": "the third field of this file's ROW KEY. Identical to `holo_chain` on every row by construction (at-rest R6); published under the key's own name so the grouping can be reconstructed from the file. Known redundancy, 2026-09-14" } ], "data_file_sha256": "fdb41b55eb429b9284a83d4eca48ce387a4813fb24c4f36fdc607b9c8b5dc920" }