Start with the biological or physical system, then choose its representation. The correct input is not always the largest available file or the longest possible sequence.
Check the service contract
Each service form documents accepted entities, formats, required fields, and limits. These vary between tools and versions. Review the current form before preparing a large dataset or reusing an example from another service.
| Input | Preserve | Check before submission |
|---|---|---|
| Protein sequence | Sequence identity, construct boundaries, mutations | Alphabet, chain count, length, and whether a header is accepted |
| DNA or RNA sequence | Correct nucleotide type and strand | Alphabet, orientation, modified bases, and intended partners |
| PDB or mmCIF structure | Chain IDs, numbering, partners, ligands, ions | Model selection, gaps, alternate positions, and tool-specific component handling |
| Small molecule | Identity, stereochemistry, charge, bonding | Accepted SMILES or coordinate format and protonation assumptions |
| Alignment or sequencing files | Headers, pairing, sequence correspondence | Tool-specific layout, encoding, and size limits |
| Periodic crystal | Cell, periodicity, composition, atom positions | Accepted CIF or POSCAR format and support for disorder |
Protein and nucleotide sequences
Keep a named copy of each submitted sequence. For complexes, represent separate entities and their copy counts according to the tool’s form. Do not join interacting chains with an artificial linker unless that linker is part of the intended construct.
Distinguish a mature protein from a precursor with a signal peptide, purification tag, or propeptide. Truncation can remove an interface or destabilize a domain; document why a construct boundary was chosen.
Use the form’s nucleotide type controls for DNA and RNA where provided. Do not silently convert modified residues or ambiguous symbols to a different identity.
Protein structures and complexes
Choose the intended biological assembly and structural model. Check for missing residues, missing side-chain atoms, duplicate chains, and alternate locations. Preserve catalytic metals and relevant cofactors unless the downstream method requires a different representation.
Residue labels used for hotspots, fixed positions, or restraints must match the actual uploaded file. A sequence index is not necessarily the same as a PDB residue number; insertion codes and unresolved segments can matter.
See Repair and validate a structure for a deliberate preparation procedure.
Small molecules
Use a representation that preserves bonding and stereochemistry. SMILES describes molecular connectivity; an SDF can also contain coordinates. A molecule name alone may not uniquely define the desired chemical form.
Specify relevant protonation, tautomer, salt, and stereochemical assumptions. Keep ligand coordinates in the receptor’s frame when the workflow depends on an existing pose. Do not treat separately generated coordinates as an already aligned complex.
Libraries and repeated inputs
Use unique, stable identifiers and validate a small subset before scaling. Check the intended pairing rule: one receptor with many ligands, matched pairs, and all combinations produce different numbers of calculations.
Preserve the manifest that maps identifiers to sequences, structures, compounds, and outputs. Review candidate expansion and compute implications before a bulk run.
Before you submit
- Confirm identity, copy counts, and required molecular partners.
- Check format and current field limits.
- Validate numbering and any selected regions.
- Inspect the input with an appropriate viewer.
- Record scientific assumptions, settings, and a recognizable job note.