Query · SMARTFlexDB

Analyze your own apo–holo pair

Upload a ligand-bound (holo) RNA structure and a ligand-free (apo) structure of the same RNA, then specify the ligand residue name. SMARTFlexDB runs the core structural analysis available for curated pairs, computed live from your two files: binding site, sequence alignment, per-residue displacement, 3D superposition + morph, secondary structure, conformational dashboard, and ligand–RNA interactions. RCSB-derived metadata and ligand descriptors are excluded.

Holo structure (ligand-bound) - mmCIF (.cif), the file containing your ligand

Apo structure (ligand-free) - mmCIF (.cif) of the same RNA, no ligand

Ligand residue name - the HET code as it appears in the holo file (e.g. 2QC). If it appears more than once, specify a unique ligand instance as LIG:CHAIN:RESSEQ (e.g. CNY:C:41), where CHAIN is the auth chain the ligand residue itself sits on.

Each file must be mmCIF and under 30 MB.

Try a worked example — load it into the form, then press Run analysis:

Each result is tagged by how its data is produced:
tool-derived produced from your coordinates by an established tool (x3dna-dssr, US-align / Kabsch, RDKit)
geometry-derived computed from coordinates
heuristic a rule-based inference — treat as a hypothesis.

Nothing is fetched from RCSB for your uploads. An uploaded result is a subset of a curated pair — no RCSB metadata and no curated BLAST identity.

What we do with your files — Scope and limitations

We locate your ligand in the holo structure and check it is a single-copy small-molecule HET sitting in an RNA pocket, superpose the apo onto the holo, and compute every panel directly from your two structures. A few things to keep in mind:

Computed live, from your coordinates. Binding-site residues, per-residue displacement, contacts, and base-pair annotations are computed locally using DSSR and Gemmi. The superposition is the same sequence-anchored Kabsch fit the rest of the site uses (a US-align fit is computed alongside and the lower-RMSD one kept; Kabsch wins on every served record). US-align is the sole source of the TM-score. tool-derived (x3dna-dssr, US-align / Kabsch)geometry-derived (gemmi distances)
No RCSB-derived metadata. Uploaded structures have no PDB accession, so the live RCSB-entry sidebar (entity counts / classification / authors) is skipped. The ligand-properties card (MW, logP, SMILES, 2D structure, etc.) is also unavailable, because those descriptors come from the ligand's RCSB chemical entry. Method, resolution, title and RNA type are still parsed from your CIF header and appear in the Overview.
The morph is interpolation, not MD. The morph and the conformational dashboard are a straight-line path between your two real endpoints — a visualization, not a trajectory. heuristic
RNA-only, single ligand. The holo must contain exactly one residue with the name you give, and it must be a real small-molecule ligand (not water / ion / a nucleotide / an amino acid). There must also be RNA within ~6 Å of it, and the apo must contain an RNA chain. apo and holo must be a closely matched RNA at comparable declared (SEQRES) length — chains differing by more than ~20% there are rejected. Unmodeled density is therefore not a rejection reason; only if one of your files declares no SEQRES at all do we fall back to its modeled nucleotide count. You will get a precise message if any check fails.
Two chains, two roles. The CHAIN you type in LIG:CHAIN:RESSEQ is the auth chain the ligand sits on — it only says which copy of the ligand to analyze. The holo chain reported in the results is the RNA chain the ligand binds. We determine it from contact distance in your file, not from what you type. It is the chain every 2D and 3D panel, the sequence alignment, the superposition and the per-residue displacement refer to. Usually they are the same strand — but in a complex or a duplex pocket a ligand can sit on one chain and bind another, and then the results will name a chain you did not type. geometry-derived

This is a temporary analysis job on the structures you provide — results are not saved to the public dataset. For how each feature is produced on the curated pairs, see About.