Close the loop between generative models and in-vitro reality. Submit candidate sequences for reviewed execution; our autonomous lab translates suitable designs into controlled wet-lab workflows and structured results.
import programmablebio as pb
client = pb.Foundry(api_key="pb_live_...")
experiment = client.experiments.create(
name="binder-screen-r04",
target="HER2_ECD",
designs="candidates.fasta",
assays=[
pb.assays.BindingKinetics(),
pb.assays.Thermostability(),
pb.assays.ExpressionYield()
]
)
# proposed standardized interface
print(experiment.status)
→ FEASIBILITY_REVIEWStructure prediction and de novo protein generation have accelerated design in the world of bits. Validation in the world of atoms remains constrained by manual handling, reagent lead times, consumable costs, and protocols that were never designed for machine-scale iteration.
candidate sequences generated
variants commonly validated
Abstract laboratory complexity into a repeatable execution layer: sequence in, evidence and provenance out.
Three integrated layers turn a candidate library into a measurable design–build–test–learn cycle.
Miniaturized workflows are designed to reduce manual transfers and reagent use while increasing the number of conditions that can be explored per run.
Move beyond binary readouts toward kinetic, stability, expression, and yield measurements that support real design decisions.
Normalize experimental output into versioned data objects built to flow back into feature stores, notebooks, and active-learning loops.
FASTA or CSV sequences with target and control specifications.
Feasibility, synthesis-rule checks, scope, and quoted execution plan.
Automated expression and instrumented assay workflows.
Measurements, raw files, QC context, and run provenance.
Programs begin with technical review. Self-serve ordering and programmatic dispatch follow as each workflow is standardized.
See how it works →The matrix below describes the standardization direction for the platform. Availability and turnaround are confirmed during feasibility review for every pilot.
| Assay class | Primary readouts | Method direction | Status |
|---|---|---|---|
| Binding kineticsInteraction quality | kon koff KD | BLI / label-free kinetic methods | Pilot review |
| ThermostabilityFolding and aggregation | Tm Tonset Tagg | Thermal shift / light-scattering methods | Planned expansion |
| Expression & yieldManufacturability signal | Soluble yield · total yield | Fluorometric / capillary methods | Pilot review |
| Enzyme kineticsFunctional performance | Vmax Km kcat | Absorbance / fluorescence traces | Custom scope |
No fixed turnaround is promised until input materials, controls, assay maturity, and capacity are reviewed.
The product direction is an authenticated interface that lets computational teams validate, dispatch, observe, and retrieve experiments without losing lineage.
Workflow-nativePlanned adapters for Nextflow and Snakemake pipelines.
Analysis-readyJSON objects designed for Pandas, Polars, and PyTorch.
Deterministic lineageConnect generation runs, synthesis lots, telemetry, and raw data.
$ pip install programmablebio
# interface preview; package not yet public
$ pb validate \
--library ./candidates.fasta \
--platform nanoliter-v2
✓ 96 sequences parsed
✓ synthesis rules checked
! 3 variants need manual review
$ pb deploy \
--target HER2_ECD \
--assay binding_kinetics
Experiment prepared for feasibility review.Generative biology needs benchmarks that connect model outputs to physical measurements. We intend to publish carefully scoped, de-identified datasets that make that gap visible and measurable.
Explore illustrative programsStart with a non-confidential description of your candidate library. We’ll assess fit, define the assay scope, and return a proposed execution plan.