Boltz-2

Predict complex structures containing proteins, nucleic acids and small molecules, with structural confidence metrics for each candidate.

API/CLI tool ID: boltz

The input fields depend on the value of inputFormat.

Accepts binding-site restraints as input.

Understanding inputs and outputs

A complex is a structure formed by multiple molecules together. In molecules input, distinguish proteins, DNA, RNA and ligands by type and chain. A colon in sequence input separates protein chains, not independent jobs.

The result table below provides structural confidence metrics. confidence_score is an overall candidate confidence score, not a binding affinity value. pTM assesses the overall structural arrangement, and ipTM assesses interfaces between chains. An empty ipTM for a single chain is not a low score. complex_plddt uses a 0–1 scale; numerical rescaling alone does not make it a performance comparison with another tool’s 0–100 pLDDT.

Input settings

SettingFormatRequiredDefaultDescription and constraints
inputFormattaskOptional"sequence"Input format (list: upload a FASTA list) — Allowed values sequence · list · molecules · yaml
sequencesequenceConditionally required—Separate chains with ':'. X residues are allowed. — Characters ACDEFGHIKLMNPQRSTVWYX:, Applies to sequence
listFilefileConditionally required—FASTA/sequence list file — Extensions txt · fa · fasta, Applies to list
moleculesmoleculesConditionally required—JSON array, Applies to molecules · list, Example [{"type":"protein","sequence":"TRPNHTIYINNLNEKIKKDELKKSLHAIFSRFGQIL","chain":"A"}]
proteinMsaModedropdownConditionally required—Protein MSA source. Only single-sequence runs are supported for now, with the possible loss in quality acknowledged. — Allowed values single-sequence, Applies to molecules, Example "single-sequence"
yamlFilefileConditionally required—Boltz YAML. Specify MSA paths and data inside the file yourself. — Extensions yaml · yml, Applies to yaml
numSamplesnumberOptional1Number of samples to generate — Minimum 1, Maximum 64
numRecyclesnumberOptional3Minimum 0, Maximum 10
stepScalenumberOptional1Lower values increase diversity (1–2 recommended)
seednumberOptional—Random seed for reproducibility (blank: random) — Minimum 0, Maximum 4294967295
versiondropdownOptional"2.2.1"Boltz code version (github.com/jwohlwend/boltz/releases) — Allowed values 2.2.1 · 1.0.0 · 0.4.0

Output metrics

ColumnDisplay nameRangeDescription
confidence_scoreOverall confidence0~1Boltz overall confidence (0–1). Not an experimental success probability or binding affinity.
complex_plddtMean pLDDT (0–1)0~1Uses Boltz's native 0–1 scale. Don't multiply by 100 to compare with other tools' pLDDT.
ptmpTM0~1—
iptmipTM0~1Inter-chain interface confidence. Undefined, and left blank, for a single chain.

See Reading results for metric definitions and interpretation. A range shows possible values; whether higher or lower is better depends on the metric.

Create plans for multiple sequences

See Preparing Boltz batches for CSV/FASTA batches of single protein chains and setting combinations. Batch support is narrower than the tool's full input contract. For plans made with a Nadoo Bio API key, a person reviews inputs, settings, and terms, then approves selected plans on the batch review page. After checking each plan's current conditions, submit through submit_batch. Use submit_plan individually if terms hashes differ. An agent cannot consent on the user's behalf.

Execution availability

Open an execution plan in your project to see available tool versions and costs. Review and approve the inputs, settings, and terms before running.

License

Code, model weights, and accompanying data have their own terms. The table summarizes terms from official sources; check your execution plan for the configuration you will use. License texts are available in Third-party notices.

ComponentLicenseCommercial useSourceNotes
CodeMITAllowed, subject to notice requirementsLicense text—
Model weightsMITAllowed, subject to notice requirementsRepository READMEThe README states that both code and weights are MIT licensed

Checked on 2026-11-01.

References

On this page