Boltz Small Molecule Explore
Budgeted search of a large SMILES library against a protein target.
Input
Boltz Small Molecule Explore webserver overview
Boltz Small Molecule Explore searches a large SMILES library against a protein target and scores only a set number of molecules from it, chosen by Boltz, instead of scoring every entry. A job returns those molecules ranked, each with a predicted complex.
ProteinIQ sends the job to the hosted Boltz API pipeline and publishes the finished table, structures and records. To score every molecule in a list of up to 1,000, use Boltz Small Molecule Screen.
Pricing
Explore costs a flat rate per scored molecule, so the price follows the Budget setting and not the size of the library. The total is rounded once for the whole job.
| Budget | Credits |
|---|---|
| 1 | 51 |
| 10 | 507 |
| 100 | 5,064 |
| 1,000 | 50,633 |
| 2,000 | 101,266 |
Library size, target size, reference ligands and filters do not change the price. The exact quote is calculated before submission. If Boltz estimates a noticeably higher cost when the job starts, the run is not started and the credits are refunded.
Inputs
| Input | Accepted formats | Limits and behavior |
|---|---|---|
Target protein | FASTA text, .fasta, .fa or .txt file, or a UniProt sequence | Required. Up to 10 chains, files up to 10 MB, one sequence each. Chain IDs are assigned A, B, C in submission order. |
Molecule library | One .csv or .tsv file | Required. Exactly one file per job, up to 50 MB, with a SMILES column. |
Reference ligand | One SMILES string, a .smi or .smiles file, or a PubChem compound | Optional. Up to 10 known binders that guide the search. |
The library file is passed to Boltz as it is, so the column names in its header must match the SMILES column and ID column settings. Each target input accepts 1 to 10 copies, and each copy becomes its own chain. Job name is an optional label for the saved run.
Settings
Search
| Parameter | Type | Default | Description |
|---|---|---|---|
Budget (budget) | integer | 100 | Molecules Boltz scores from the library, from 1 to 2,000. Each one is billed. |
SMILES column (smiles_column) | string | smiles | Library column that holds SMILES. |
ID column (id_column) | string | optional | Library column used as the molecule ID. Without it, Boltz names the results itself. |
Pocket residues (pocket_residues) | string | optional | Binding-pocket residues as 1-based positions, for example A:45,48-52. Without a chain prefix the first target chain is used. |
Molecule filters
Explore applies no structural alert filter unless one is set here.
| Parameter | Type | Default | Description |
|---|---|---|---|
Lipinski filter (lipinski_filter) | boolean | false | Removes molecules that break Lipinski's rule of five. |
Maximum molecular weight (lipinski_max_mw) | number | 500 | Largest molecular weight kept, in Da. Applies when the Lipinski filter is on. |
Maximum LogP (lipinski_max_logp) | number | 5 | Largest LogP kept. Applies when the Lipinski filter is on. |
Maximum H-bond donors (lipinski_max_hbd) | integer | 5 | Most hydrogen-bond donors kept. Applies when the Lipinski filter is on. |
Maximum H-bond acceptors (lipinski_max_hba) | integer | 10 | Most hydrogen-bond acceptors kept. Applies when the Lipinski filter is on. |
Allow one violation (lipinski_allow_single_violation) | boolean | false | Keeps molecules that break at most one Lipinski rule. |
Additional alert catalog (alert_catalog) | enum | none | One RDKit structural-alert catalog: PAINS, PAINS A, PAINS B, PAINS C, BRENK, NIH, or a CHEMBL catalog. |
Excluded SMARTS (excluded_smarts) | string | optional | SMARTS patterns to exclude, one per line or comma-separated. |
Outputs
Viewer shows the predicted complexes. Results lists the scored molecules. Files holds the downloads.
| Download | Contents |
|---|---|
boltz_sm_explore_0001_<name>.cif onwards | Predicted complex for each scored molecule, numbered by rank. |
boltz_sm_explore_results.csv | The results table. |
boltz_sm_explore_results.json | Boltz's run and per-result records. |
The scored molecules, the CSV table and the predicted complexes can be passed to other tools in workflows.
Understanding results
Each row carries a rank, the molecule ID, its SMILES, Boltz's own metric columns, any warnings Boltz attached and the matching structure file. Ranks follow the order Boltz returns, so rank 1 is its top-scoring entry. Nested metrics are flattened into columns joined by underscores.
The results cover only the molecules Boltz chose to score, so a library entry missing from the table was not evaluated rather than scored poorly. Scores order candidates within one run and are not measured affinities.
Related tools
Boltz Small Molecule Screen
Rank your SMILES library against a protein target.
AF2BIND
Predict protein ligand-binding residues from AlphaFold2 pair representations
Boltz ADME
LogD, permeability and solubility predictions for small molecules.
SPRINT
Rank compound libraries with SPRINT protein and ligand co-embeddings.
BoltzMol-1
Small-molecule design against a protein target from purchasable chemical space.
CatPred
Predict enzyme kcat, Km and Ki from sequences and SMILES, with uncertainty.
DLKcat
Predict enzyme kcat values from protein sequences and substrate structures or names.
PandaDock
Physics-based molecular docking with CPU-optimized hierarchical algorithms
PLIP
Profile protein-ligand interactions from a PDB complex structure.
PoseBusters
Validate generated and docked ligand poses with structure, chemistry, and reference-pose checks.