openprotein.fold#

Parameters:: job (Job) – The fold job whose results are to be retrieved.
Returns:: An instance of FoldResultFuture
Return type:: FoldResultFuture

Models#

class openprotein.fold.ProtenixModel(session, model_id, metadata=None)[source]#

Class providing inference endpoints for Protenix structure prediction.

fold(sequences, diffusion_samples=1, num_recycles=10, num_steps=200, templates=None, **_)[source]#

Request structure prediction with Protenix.

Parameters:

sequences (Sequence[Complex | Protein | str | bytes] | MSAFuture) – List of protein complexes to include in folded output. Protein objects must be tagged with an msa, which can be a Protein.single_sequence_mode for single sequence mode. Alternatively, supply an MSAFuture to use all query sequences as a multimer.
diffusion_samples (int) – Number of diffusion samples to use
num_recycles (int) – Number of recycling steps to use
num_steps (int) – Number of sampling steps to use
templates (list[Protein | Complex | Template] | None = None) – List of templates to use for structure prediction.

Returns:

Future for the folding results.

Return type:

FoldResultFuture

class openprotein.fold.Boltz2Model(session, model_id, metadata=None)[source]#

Class providing inference endpoints for Boltz-2 structure prediction model which jointly models complex structures and binding affinities.

fold(sequences, diffusion_samples=1, num_recycles=3, num_steps=200, step_scale=1.638, use_potentials=False, constraints=None, templates=None, properties=None, method=None, **_)[source]#

Request structure prediction with Boltz-2 model.

Parameters:

sequences (Sequence[Complex | Protein | str | bytes] | MSAFuture) – List of protein sequences to include in folded output. Protein objects must be tagged with an msa, which can be a Protein.single_sequence_mode for single sequence mode. Alternatively, supply an MSAFuture to use all query sequences as a multimer.
diffusion_samples (int) – Number of diffusion samples to use
num_recycles (int) – Number of recycling steps to use
num_steps (int) – Number of sampling steps to use
step_scale (float) – Scaling factor for diffusion steps.
use_potentials (bool = False.) – Whether or not to use potentials.
constraints (list[dict] | None = None) – List of constraints.
templates (list[Protein | Complex | Template] | None = None) – List of templates to use for structure prediction.
properties (list[dict] | None = None) – List of additional properties to predict. Should match the BoltzProperties
method (str | None) – The experimental method or supervision source used for the prediction. Defults to None. Supported values (case-insensitive) include: ‘MD’, ‘X-RAY DIFFRACTION’, ‘ELECTRON MICROSCOPY’, ‘SOLUTION NMR’, ‘SOLID-STATE NMR’, ‘NEUTRON DIFFRACTION’, ‘ELECTRON CRYSTALLOGRAPHY’, ‘FIBER DIFFRACTION’, ‘POWDER DIFFRACTION’, ‘INFRARED SPECTROSCOPY’, ‘FLUORESCENCE TRANSFER’, ‘EPR’, ‘THEORETICAL MODEL’, ‘SOLUTION SCATTERING’, ‘OTHER’, ‘AFDB’, ‘BOLTZ-1’. View the documentation on Boltz for upstream details.

Returns:

Future for the folding result.

Return type:

FoldResultFuture

class openprotein.fold.Boltz1xModel(session, model_id, metadata=None)[source]#

Class providing inference endpoints for Boltz-1x open-source structure prediction model, which adds the use of inference potentials to improve performance.

fold(sequences, diffusion_samples=1, num_recycles=3, num_steps=200, step_scale=1.638, use_potentials=True, constraints=None, **_)[source]#

Request structure prediction with Boltz-1x model. Uses potentials with Boltz-1 model.

Parameters:

sequences (Sequence[Complex | Protein | str | bytes] | MSAFuture) – List of protein sequences to include in folded output. Protein objects must be tagged with an msa, which can be a Protein.single_sequence_mode for single sequence mode. Alternatively, supply an MSAFuture to use all query sequences as a multimer.
diffusion_samples (int) – Number of diffusion samples to use
num_recycles (int) – Number of recycling steps to use
num_steps (int) – Number of sampling steps to use
step_scale (float) – Scaling factor for diffusion steps.
constraints (Optional[List[dict]]) – List of constraints.

Returns:

Future for the folding complex result.

Return type:

FoldResultFuture

class openprotein.fold.Boltz1Model(session, model_id, metadata=None)[source]#

Class providing inference endpoints for Boltz-1 open-source structure prediction model.

fold(sequences, diffusion_samples=1, num_recycles=3, num_steps=200, step_scale=1.638, use_potentials=False, constraints=None, **_)[source]#

Request structure prediction with Boltz-1 model.

Parameters:

sequences (Sequence[Complex | Protein | str | bytes] | MSAFuture) – List of protein sequences to include in folded output. Protein objects must be tagged with an msa, which can be a Protein.single_sequence_mode for single sequence mode. Alternatively, supply an MSAFuture to use all query sequences as a multimer.
diffusion_samples (int) – Number of diffusion samples to use
num_recycles (int) – Number of recycling steps to use
num_steps (int) – Number of sampling steps to use
step_scale (float) – Scaling factor for diffusion steps.
use_potentials (bool = False.) – Whether or not to use potentials.
constraints (Optional[List[dict]]) – List of constraints.

Returns:

Future for the folding complex result.

Return type:

FoldResultFuture

class openprotein.fold.RosettaFold3Model(session, model_id, metadata=None)[source]#

Class providing inference endpoints for RosettaFold-3 structure prediction model.

fold(sequences, diffusion_samples=1, num_recycles=10, num_steps=50, **kwargs)[source]#

Request structure prediction with RosettaFold-3 model.

Parameters:

sequences (list[Complex | Protein | str | bytes] | MSAFuture,) – List of protein sequences to include in folded output. Protein objects must be tagged with an msa, which can be a Protein.single_sequence_mode for single sequence mode. Alternatively, supply an MSAFuture to use all query sequences as a multimer.
diffusion_samples (int) – Number of diffusion samples to use
num_recycles (int) – Number of recycling steps to use
num_steps (int) – Number of sampling steps to use

Returns:

Future for the folding results.

Return type:

FoldResultFuture

class openprotein.fold.AlphaFold2Model(session, model_id, metadata=None)[source]#

Class providing inference endpoints for AlphaFold2 structure prediction models, based on the implementation by ColabFold.

fold(sequences, num_recycles=None, num_models=1, num_relax=0, **kwargs)[source]#

Post sequences to alphafold model.

Parameters:

sequences (Sequence[Complex | Protein | str | bytes] | MSAFuture) – List of protein sequences to include in folded output. Protein objects must be tagged with an msa, which can be a Protein.single_sequence_mode for single sequence mode. Alternatively, supply an MSAFuture to use all query sequences as a multimer.
num_recycles (int) – number of times to recycle models
num_models (int) – number of models to train - best model will be used
num_relax (int) – maximum number of iterations for relax

Returns:

job

Return type:

Job

class openprotein.fold.ESMFoldModel(session, model_id, metadata=None)[source]#

Class providing inference endpoints for Facebook’s ESMFold structure prediction models.

fold(sequences, num_recycles=None, **_)[source]#

Fold sequences using this model.

Parameters:

sequences (Sequence[Complex | Protein | str | bytes]) – sequences to fold
num_recycles (int | None) – number of times to recycle models

Return type:

FoldResultFuture

class openprotein.fold.MiniFoldModel(session, model_id, metadata=None)[source]#

Class providing inference endpoints for MiniFold.

fold(sequences, num_recycles=None, **_)[source]#

Fold sequences using this model.

Parameters:

sequences (Sequence[bytes | str]) – sequences to fold
num_recycles (int | None) – number of times to recycle models

Return type:

FoldResultFuture

Results#

class openprotein.fold.FoldResultFuture(session, job=None, metadata=None, sequences=None, complexes=None, max_workers=10)[source]#

Fold results represented as a future.

job#

The fold job associated with this future.

Type:: FoldJob

property sequences: list[bytes]#

Get the sequences submitted for the fold request.

Returns:: List of sequences.
Return type:: list[bytes]

property complexes: list[Complex]#

Get the molecular complexes submitted for the fold request.

Returns:: List of complexes.
Return type:: list[Complex]

property id#

Get the ID of the fold request.

Returns:: Fold job ID.
Return type:: str

property metadata: FoldMetadata#: The fold metadata.

property model_id: str#: The fold model used.

get_item(k: int, key: None = None) → Structure[source]#

get_item(k: int, key: Literal['pae', 'pde', 'plddt', 'ptm', 'ipae'] | None = None) → ndarray

get_item(k: int, key: Literal['affinity']) → BoltzAffinity

get_item(k: int, key: Literal['confidence']) → list[BoltzConfidence]

get_item(k: int, key: Literal['score', 'metrics'] | None = None) → DataFrame

Get fold results for a specified sequence.

Parameters:: sequence (bytes) – Sequence to fetch results for.
Returns:: Complex containing the folded structure.
Return type:: Complex

stream(key: None = None) → Iterator[Structure][source]#

stream(key: Literal['pae', 'pde', 'plddt', 'ptm', 'ipae'] | None = None) → Iterator[ndarray]

stream(key: Literal['affinity']) → Iterator[BoltzAffinity]

stream(key: Literal['confidence']) → Iterator[list[BoltzConfidence]]

stream(key: Literal['score', 'metrics'] | None = None) → Iterator[DataFrame]

Retrieve results for this job as a stream.

Returns:: A generator that yields (key, value) tuples.
Return type:: Generator

get_pae()[source]#

Get the Predicted Aligned Error (PAE) matrix for all outputs.

Returns:: PAE matrix.
Return type:: list[np.ndarray]
Raises:: AttributeError – If PAE is not supported for the model.

property args: dict[str, Any]#: The registered job arguments.

cancelled()#

Check if the job has been cancelled.

Returns:: True if the job is cancelled, False otherwise.
Return type:: bool

property created_date: datetime#: The creation timestamp of the job.

done()#

Check if the job has completed.

Returns:: True if the job is done, False otherwise.
Return type:: bool

property end_date: datetime | None#: The end timestamp of the job.

get_pde()[source]#

Get the Predicted Distance Error (PDE) matrix.

Returns:: PDE matrix.
Return type:: list[np.ndarray]
Raises:: AttributeError – If PDE is not supported for the model.

property job_id: str#: The unique identifier of the job.

property job_type: str#: The type of the job.

property progress_counter: int#: The progress counter of the job.

refresh()#: Refresh the job status and internal job object.

property start_date: datetime | None#: The start timestamp of the job.

property status: JobStatus#: The current status of the job.

wait(interval=5, timeout=None, verbose=False)#

Wait for the job to complete, then fetch results.

Parameters:

interval (int, optional) – Time in seconds between polling. Defaults to config.POLLING_INTERVAL.
timeout (int | None, optional) – Maximum time in seconds to wait. Defaults to None.
verbose (bool, optional) – Verbosity flag. Defaults to False.

Returns:

The results of the job.

Return type:

Any

wait_until_done(interval=5, timeout=None, verbose=False)#

Wait for the job to complete.

Parameters:

interval (float, optional) – Time in seconds between polling. Defaults to config.POLLING_INTERVAL.
timeout (int, optional) – Maximum time in seconds to wait. Defaults to None.
verbose (bool, optional) – Verbosity flag. Defaults to False.

Returns:

True if the job completed successfully.

Return type:

bool

Notes

This method does not fetch the job results, unlike wait().

get_plddt()[source]#

Get the Predicted Local Distance Difference Test (pLDDT) scores.

Returns:: pLDDT scores.
Return type:: list[np.ndarray]
Raises:: AttributeError – If pLDDT is not supported for the model.

get_ptm()[source]#

Get the Predicted TM (pTM) scores.

Returns:: pTM scores.
Return type:: list[np.ndarray]
Raises:: AttributeError – If pTM is not supported for the model.

get_ipae()[source]#

Get the interface PAE (iPAE) — a synthetic scalar per unit derived from pae and the per-unit protein-chain layout. Returns one shape-(1,) array per fold output.

Returns:: iPAE scalars (one (1,) array per unit).
Return type:: list[np.ndarray]
Raises:: AttributeError – If iPAE is not supported for the model.

get_score()[source]#

Get the predicted scores.

Returns:: Structure prediction scores.
Return type:: list[pd.DataFrame]
Raises:: AttributeError – If score is not supported for the model.

get_metrics()[source]#

Get the predicted metrics.

Returns:: Structure prediction metrics.
Return type:: list[pd.DataFrame]
Raises:: AttributeError – If metrics is not supported for the model.

get_confidence()[source]#

Retrieve the confidences of the structure prediction.

Note

This is currently supported for Boltz models and Protenix.

Returns:: List of list of confidence objects (model-specific schema).
Return type:: list[list[BoltzConfidence]] | list[list[ProtenixConfidence]]
Raises:: AttributeError – If confidence is not supported for the model.

get_affinity()[source]#

Retrieve the predicted binding affinities.

Note

This is only currently supported for Boltz models.

Returns:: BoltzAffinity object containing the predicted affinities.
Return type:: list[list[BoltzAffinity]]
Raises:: AttributeError – If affinity is not supported for the model.

get_pae_batch()[source]#

Get the Predicted Aligned Error (PAE) for every unit as a single stacked np.ndarray of shape [N, ...]. Per-unit arrays that differ in size are NaN-padded to the per-axis max shape.

get_pde_batch()[source]#

Like get_pae_batch but for PDE. Shape [N, ...], NaN-padded.

get_plddt_batch()[source]#

Like get_pae_batch but for pLDDT. Shape [N, ...], NaN-padded.

get_ptm_batch()[source]#

Like get_pae_batch but for pTM. Shape [N, ...], NaN-padded.

get_ipae_batch()[source]#

Get iPAE for every unit as a single np.ndarray of shape [N]. Units whose per-unit iPAE could not be computed appear as NaN.

get_confidence_batch()[source]#

Retrieve per-unit confidence objects in a single HTTP call.

Returns a length-N list; each entry is that unit’s parsed confidence list, or None if the server could not fetch the result for that unit.

get_affinity_batch()[source]#

Retrieve per-unit affinity objects in a single HTTP call.

Returns a length-N list; each entry is a BoltzAffinity or None if the server could not fetch that unit’s result.