SMARTFlexDB

A ligand-bound (holo) RNA binding site matched to a closely matched ligand-free (apo) structure of the same RNA (≥90% sequence identity over ≥90% coverage), where the corresponding pocket is unoccupied. Comparing the two states shows the conformational differences associated with ligand binding.

Three counts of the same release, at three levels of detail:

198 = unique RNA–ligand entries, each a distinct (holo structure · RNA chain · ligand · apo) combination after merging crystallographic copies of the same complex — what the gallery shows as cards.

266 = binding sites (pockets), each paired with its default apo structure. A pocket is one ligand instance in one holo structure — formally the combination (holo PDB · ligand · the chain and residue number that ligand copy sits on), which is unique across all 266. So a crystal containing the same complex several times contributes one record per copy: 8D5L, for example, has the ligand QIJ bound four times (chains A, B, C, D) and so appears as four pockets. 55 holo structures contribute more than one. This is the level GET /api/pairs returns.

497 = apo–holo comparisons: the 266 default pairings plus every alternative apo structure offered for the same pocket (231 of them). Many pockets have more than one experimental apo structure available, and each alternative is its own comparison — this is the level the bulk TSV lists, one row per comparison. Together they span 304 PDB structures (193 holo + 111 apo).

Two numbers are published for every pair, because they answer different questions:

Local sequence identity scores only the region the two sequences align over, so a mismatch in a terminal overhang is trimmed out rather than counted. Every served pair is ≥90% on this measure — "closely matched RNA", not necessarily identical.

Global sequence identity scores the entire declared (SEQRES) sequence, so every mismatch and every gap counts, terminal overhangs included. It is therefore never higher than the local value, and the gap between the two tells you how much of the difference sits outside the aligned core.

For a multi-chain pocket the published value is the ligand's own (auth) chain's — the chain the PDB entry assigns the ligand to, the same one the pocket identifier carries. Every chain's own value is shown on the pair page beside that chain's alignment.

No. The morph is a straight-line interpolation between the two superposed experimental endpoints, for visualization only — it is not an MD or kinetic pathway.

Playback eases in and out, so the two endpoint states are held long enough to read. That is a display choice and carries no kinetic meaning: the frames lie on the same straight line either way, and the easing changes only how they are spaced in time.

Per-selection CSV from the Database toolbar, per-structure files from each pair page, cluster assignments from Clustering, the whole dataset as flat files and a structure bundle from the API page, or programmatically via the REST API.