Molecular construction

The construction API prepares starting geometries; it does not predict a stable conformation or a metal’s preferred coordination geometry. Coordinates are angstroms and rotations are degrees.

molecular_builder.rotate_bond(structure, [i,j], angle, frozen=...) cuts the selected bond and rotates the smaller connected component around i→j. Ties move the j-side, matching exact parameter edits. Ring bonds and movements of frozen atoms refuse without modifying the source.

coordination_builder.coordinate fills open sites of linear, trigonal planar, tetrahedral, square planar, trigonal bipyramidal, square pyramidal and octahedral templates. Choose the metal–donor distance explicitly. Existing ligands orient the template and remain unchanged; optionally replace directly bonded hydrogen atoms. Water, ammonia, chloride, carbon monoxide, pyridine, cyanide, ethylenediamine and oxalate are available. Chelates are placed rigidly: a mismatch between their donor separation and the selected sites refuses instead of stretching bonds. Template placement is not a metal force-field calculation.

molecular_builder.conformers(molecule, count, seed) uses RDKit ETKDGv3, one thread and a fixed seed, followed by MMFF94 or UFF when fully parameterised. Reproducibility is scoped to the recorded RDKit version. It retains optimizer convergence flags, ranks energies in kcal/mol, and writes all conformers as selectable structures and trajectory frames. Unparameterised molecules refuse energy ranking. The current-structure adapter requires complete hydrogens and valences; it does not infer missing atom-local formal charges.

peptide_builder.peptide(sequence, preset, phi, psi) supplies standard L-amino acid topology and imposes alpha-helix (-57,-47), sheet (-135,135), or custom phi/psi angles after embedding. Peptide omega is trans. Non-proline chains use rigid torsion edits. Internal proline uses constrained MMFF94 cleanup, with independently measured torsions within 0.1 degree, ring-angle changes at most 10 degrees, peptide-nitrogen angle sum at least 350 degrees, preserved specified stereocentres and no heavy-atom clash below 1.8 Å beyond two bonds. An infeasible requested angle refuses without returning a geometry; proline is not excluded categorically. These APIs require the optional [smiles] extra where RDKit is involved.

Methods: RDKit conformer documentation, Riniker and Landrum, ETKDG, and RDKit FASTA topology API.

Builder panel

The Molecular builder card provides typed or slider bond rotation (slider release commits one undo step; the bond stays selected for further steps), coordination-site filling on a selected metal, sequence construction, current-molecule conformers and a searchable library of 42 small fragments with generated 2D thumbnails. Open starts a new molecule; Attach replaces one explicit hydrogen at the selected atom and one at the fragment’s first atom. Atom/bond edits preserve explicit connectivity. Unfreeze atoms before topology-changing operations; rigid rotation can retain frozen atoms on its axis. All builder edits stop automatic optimization.

Sequences use the RDKit FASTA topology and retain atom/residue identities. Choose single (the compatible default) or duplex for DNA/RNA. Single strands use the backbone construction below; duplexes add the antiparallel reverse complement as a distinct chain. Neither mode predicts secondary structure. Local heavy-atom templates are CC0 wwPDB coordinates from 1BNA (A6 adenine, A9 cytosine, A2 guanine, B20 thymine) and the 1EHZ RNA stem (A5 adenine, A2 cytosine, A3 guanine, A6 uracil). Hydrogens are added from explicit topology. DNA epsilon/zeta/alpha values (-178.3/-96.8/-61 degrees) follow the BI means in Svozil et al. (2008), Table 3; RNA values (-150/-73/-65 degrees) follow Schneider et al. (2004), Table 2. Local sugar conformations begin from the observed templates. Joining uses a 1.607 Å O3′–P bond, 120° C3′–O3′–P angle, 104° O3′–P–O5′ angle and tetrahedral nonbridging phosphate oxygens at 1.48 Å. A staged constrained MMFF94 cleanup removes base clashes while retaining alpha/epsilon/zeta and each template delta torsion within 0.1 degree. Other local geometry may relax. The result refuses if optimization fails, a specified stereocentre changes, a heavy bond falls outside 1.1–1.85 Å, or nonbonded heavy atoms separated by more than two bonds remain closer than 1.8 Å. These are construction safeguards, not an equilibrium-structure claim. The resulting geometries still need inspection before calculation.

Complementary duplexes

The entered sequence is chain A in the 5′→3′ direction. Chain B is its reverse complement, also numbered 5′→3′: A_i pairs with B_(n+1−i). The two chains have separate phosphodiester topology and no covalent bonds between them. Only unambiguous A/C/G/T (DNA) or A/C/G/U (RNA), up to 100 base pairs, are accepted. Duplex cleanup runs in the background and opens a new structure when accepted. The default RDKit FASTA protonation is retained; choose the required protonation and charge before a subsequent calculation.

Duplexes repeat an experimentally derived rigid helical step, fitted jointly to both strands: 1BNA A5/B20 → A6/B19 for B-family DNA (3.129153 Å helical rise, 37.444813° twist); 1EHZ A3/A70 → A4/A69 for A-family RNA (3.005436 Å, 32.778421°). These particular measured steps are not universal ideal B/A parameters. Sugar delta conformations remain those of the nucleotide templates. The matrix rebuilding formulation is described by Lu and Olson (2003); no 3DNA software or parameters distributed under its licence are included.

Watson–Crick base-pair coordinates come separately from 1BNA A5/B20 (AT) and A4/B21 (GC), or 1EHZ A5/A68 (AU) and A3/A70 (GC). They are aligned to the common pair frame. Each base is rigidly rotated to a common pair plane while preserving all its internal distances: sequence-specific experimental buckle and propeller cannot safely be repeated at arbitrary sequence boundaries. A proper half-turn exchanges purine/pyrimidine strand assignments. The JSON resource retains the original CC0 coordinates, residue identifiers, and fitted rotation/translation so the construction is reproducible.

Base heavy atoms remain fixed during MMFF94 cleanup; only backbone/hydrogen coordinates relax, with interstrand interactions enabled. The same convergence, 0.1° sugar-delta, heavy-bond, heavy-clash and stereochemistry checks used by the single-strand builder are retained. Each intended AT/AU pair must have two and GC three donor/acceptor contacts, at 2.4–3.5 Å with D–H···A angles at least 120°. Moved fixed bases or failed checks return no accepted structure. Recorded builder_pairing, builder_complement and builder_helix_step provenance follows the structure into QVF. These are inspectable starting geometries; this method does not predict solvent, ions, sequence-dependent helix parameters, or stability.

Conformer generation retains all requested frames and records energy units, force field, seed and individual minimizer convergence. Select a structure section or trajectory frame to inspect the energy-ranked alternatives. Conformer energies are force-field energies, not quantum results.