About · SMARTFlexDB

About SMARTFlexDB

SMARTFlexDB pairs ligand-bound (holo) RNA structures with ligand-free (apo) structures of the same RNA, so you can see how the RNA and its binding pocket change. Each entry is one holo structure paired with one apo structure.

Getting started Example queries Example entry How entries are selected FAQ Usage Contact License

Getting started

Browse the Database. Each card is one entry: one holo structure paired with one apo structure (a holo structure with several apo structures has one entry for each).
- Filter with the facets on the left (identity, composition, functional type, method, pocket motif, ligand molecular weight, RNA length) or search by PDB ID, ligand or title. The page opens on the whole release; Reset filters brings it back.
- To find large or small changes, Sort by TM-score or RMSD, or click a bar on Statistics.
- Table lists one row per entry; CSV downloads a ZIP with those rows and the pocket-level comparisons that meet the filters.
Holo structures (the toolbar button next to Entries) shows one card per holo PDB structure, with its number of apo structures and of related holo structures (those sharing an apo structure). Click a card for a panel listing both, always over the whole release; click an apo structure to open that entry, or View to move to a related holo structure. The Local and Global identity facets apply to the Entries view only. This view sorts by RNA length, binding-site nucleotides, ligand or PDB ID: TM-score and RMSD belong to one entry.
Open a pair. Click an entry card, or the holo PDB ID in a table row. The pair page shows the 3D superposition and the apo→holo morph (an entry without a morph says so), both 2D structures, the sequence alignment, per-residue displacement, base-pair changes, ligand–RNA interactions and downloads. If the holo structure has several binding pockets, pick one in the Binding Pocket menu next to the title.
Go deeper. Clustering: sequence, structure and pocket clusters for redundancy-aware train/test splits. Multi-ligand pocket groups: one RNA pocket bound by different ligands. Query: the same core analysis on your own apo and holo structures.

Interpretation note: Observed apo–holo conformational differences should not automatically be interpreted as resulting from ligand binding, as they may also reflect intrinsic RNA flexibility, differences in nucleotide modifications, or experimental conditions. Please interpret individual comparisons cautiously and on a case-by-case basis.

Try an example query

Each chip opens the Database with that filter applied.

SAM riboswitch Riboswitches Protein synthesis Viral RNA Aptamers Junction binding sites Internal-loop pockets X-ray structures

Example entry

overlay

Example · the adenine riboswitch

The Vibrio vulnificus adenine riboswitch aptamer bound to adenine (5SWE) and ligand-free (5E54). Both come from one X-ray free-electron laser (XFEL) study in which adenine was mixed into ligand-free crystals and the RNA switched to its bound form inside the crystal (Stagno et al., Nature 2017).

Open this entry →

How entries are selected

A holo and an apo structure form an entry only if all of these hold:

A small-molecule ligand. The holo structure binds a carbon-containing molecule made only of C, H, N, O, P, S, F, Cl, Br, B, Si, Se and Co, other than water or part of the RNA chain. There is no size limit, so buffer molecules such as glycerol can count.
A ligand-free apo. The apo structure contains no such molecule anywhere; water and ions may be present.
RNA only. Neither structure contains a protein or DNA chain.
The same RNA. Every RNA chain of the binding site (nucleotides with any heavy atom within 8 Å of a ligand heavy atom) has its own apo chain with ≥90% identity over ≥90% of both chains (BLASTN).
The same pocket. Every binding-site residue has the same base in the apo, and the apo models at least part of the site.
The same RNA molecules. Holo and apo have the same number of RNA strands, and the apo has no extra RNA molecule (such as an exon) that the holo lacks.

What the pair page shows, and where it comes from

Tags: deposited metadata read as is from the PDB entry or NAKB · tool-derived computed by a published tool (DSSR, US-align, PLIP, RDKit) · geometry-derived computed here from the deposited coordinates and sequences (gemmi, NumPy, Biopython).

Feature	What you see	How it is made
Basic Information	Overview numbers (next rows), each structure's PDB details, ligand properties	RCSB Data API (read live) and Chemical Component Dictionary; RDKit for ligand properties deposited metadatatool-derived
Global / pocket RMSD	apo→holo C3′ RMSD over all binding-site chains (global) or over the residues within 8 Å of the ligand (pocket)	after the superposition below geometry-derived
TM-score	fold similarity of the ligand's host chain, normalised by the holo length. Under ~30 nt it tracks chain length, so a low value there does not mean a different fold	US-align tool-derived
Radius of gyration	size of the fitted chains, apo → holo, over the residues modelled in both	C3′ atoms, NumPy geometry-derived
Ligand burial	share of the ligand surface covered by RNA in the holo complex	Shrake–Rupley surface area, Biopython geometry-derived
Functional type	the most specific functional class of the host chain; Unannotated when NAKB gives none	NAKB deposited metadata
Sequence alignment	apo and holo lined up residue by residue; faded letters are declared but not modelled	Needleman–Wunsch on the declared (SEQRES) sequences geometry-derived
Per-nucleotide displacement	how far each nucleotide's C3′ atom moved; one chart per binding-site chain, all in the same frame	distance after the superposition below geometry-derived
3D superposition + morph	the apo overlaid on the holo, and the morph from apo to holo (an entry without a morph says so)	one rigid Kabsch fit on the paired C3′ atoms of all binding-site chains (residues paired by sequence, one apo chain per holo chain). The morph is a straight-line interpolation for display, not a simulation geometry-derived
Secondary structure (2D)	apo and holo base pairing; click a base to find it in 3D · what DSSR extracts ↓	DSSR tool-derived
Conformational dashboard	one cell per holo nucleotide: displacement colour, binding-site marker, motif band	displacement + DSSR motifs geometry-derivedtool-derived
Base-pair rewiring	base pairs formed, disrupted or conserved between apo and holo	DSSR pairs of both structures, matched through the sequence alignment tool-derived
Ligand interactions (3D) · Interaction data	H-bonds, π-stacking, halogen bonds, ligand–metal coordination and residue distances; tables of base pairs, interactions, H-bonds, rewiring and motifs	H-bonds: DSSR (PLIP when DSSR cannot read the ligand). π-stacking, halogen bonds, metal coordination: PLIP. Distances: gemmi tool-derivedgeometry-derived

What DSSR extracts

DSSR reads the coordinates of each apo and holo structure (nothing is predicted from sequence) and reports:

2D layout

the sequence and dot-bracket of each chain → the Secondary structure (2D) diagram.

Base pairs

every pair, canonical and non-canonical, with its Leontis–Westhof and Saenger class → the Base pairing table and Base-pair rewiring.

Structural motifs

stems, hairpin, internal and bulge loops, junctions and single strands → the Structural motifs table and the motif colouring.

Reading the Leontis–Westhof code (cWW, tHS, …)

Each base has three edges: Watson–Crick, Hoogsteen and Sugar. A pair is named by its glycosidic orientation (cis or trans) and the two edges in contact, giving 12 families. cWW is the standard A–U, G–C or G–U pair; tHS is a common non-canonical pair. Non-canonical pairs, which dot-bracket cannot show, still appear in the Base pairing table.

Frequently asked questions

What is an apo–holo entry?

One holo structure (RNA bound to a small-molecule ligand) paired with one ligand-free apo structure of the same RNA. Main rules (full list under How entries are selected):

≥90% sequence identity (BLASTN) over ≥90% of both chains;
the same base at every binding-site residue (within 8 Å of the ligand);
no protein or DNA in either structure.

A holo structure paired with several apo structures has one entry for each; all count the same.

What do 1,571, 806, 271 and 443 mean?

Four ways of counting the same release:

Count	Unit	Where you see it
806	entries — one holo structure paired with one apo structure	Database Entries cards; table rows and the entry-level CSV in either view; Statistics charts of apo→holo change and identity
271	holo structures — one holo PDB structure, with all its binding pockets and ligands	Database Holo structures cards; Statistics charts of holo properties; the Clustering map
443	binding pockets — one ligand bound at one site of a holo structure	the pair page’s Binding Pocket menu; `GET /api/pairs`
1,571	pocket-level comparisons — one binding pocket against one apo structure; also the rows of the release file	the Full release file `smartflexdb_pairs.tsv` and the pocket-level CSV of a Database download; Statistics distributions

Entries. A holo structure is paired with 1 to 22 apo structures (7QP2 and 7QSH with 22), 195 distinct apo structures in all.
Holo structures. 48 bind more than one ligand, and their cards list them all (1O9M: 42B, BDG, P24, PA2); 155 distinct ligands in all. 103 hold more than one binding pocket.
Binding pockets. Identified by holo PDB, ligand, and the ligand’s chain and residue number, so one holo structure can hold several, even of one ligand (2ET5 has four RIO pockets). Each is measured on its own coordinates.
Release rows. The release file has one row per binding pocket and apo structure (1,571 rows, 24 columns). Whole-structure values such as global RMSD and TM-score are shared by binding pockets on the same RNA strand. Columns are described on the API page.

How is sequence identity defined?

Each entry has two, both comparing the holo RNA with the apo RNA:

Local identity	BLASTN (BLAST+) identity inside the aligned region only, so unmatched ends are left out. Every entry has ≥90%.
Global identity	Needleman–Wunsch alignment (1970) of the full declared (SEQRES) sequences: every mismatch and gap counts, overhanging ends included. Unmodeled residues do not lower it. Rounded to whole percent, so 100 means ≥99.5%.

For a binding site on several chains, the published value is the ligand’s own chain’s (the chain the PDB entry assigns the ligand to), shown beside that chain’s alignment on the pair page.

Which chain is the superposition fitted on?

On all RNA chains of the binding site together (each chain with a heavy atom within 8 Å of the ligand); no chain is the reference. Residues are paired by sequence alignment, each holo chain with its own apo chain, and the one rigid transform that minimises the C3′ RMSD over them (Kabsch fit) is applied to the whole apo structure.

For a multi-chain site, each chain gets two RMSDs: in the shared fit, and fitted alone. A large gap means the chains moved relative to each other.

Is the morph a molecular-dynamics trajectory?

No. It is a straight-line interpolation between the two superposed experimental structures, for visualization only. Playback slows near the two ends so they stay readable; that is a display choice with no kinetic meaning.

How do I download the data?

Your selection: CSV on the Database toolbar, a ZIP of two files: one row per entry, and one row per pocket-level comparison that meets the filters.
The whole release: smartflexdb_pairs.tsv and its .meta.json on the API page.
One binding pocket: Download pair (ZIP) on its pair page, or single files in that page’s Downloads section.
Clusters: the CSV on Clustering.
From code: the REST API.

Usage

Loading usage statistics…

Contact & feedback

Questions, bug reports, corrections, or an apo–holo pair we missed? Email yanjun.li@ufl.edu and jiang.shiyu@ufl.edu. We respond to data corrections promptly. To analyze your own apo and holo structures, use Query.

License & data sources

The SMARTFlexDB derived data (pair list, descriptors, alignments) are released under CC BY-NC 4.0. Atomic coordinates come from the RCSB PDB and remain under its terms.