Alambiq

Alambiq is software that runs computational methods for biotechnology as capabilities. A campaign, written in YAML or Python, lists steps, and each step runs one capability. A capability that runs scientific software does so in a pinned engine environment, and one that needs models or databases declares them as pinned resources. A report is one HTML file; alambiq run --report PATH writes it. The terms are defined in Concepts.

Version 0.1.0. The code, documentation and configuration written for Alambiq are Apache-2.0. Optional software, models and datasets are licensed under their own terms (LICENSING.md).

Capabilities

Capabilities are grouped in families (alambiq capability list shows the family of each):

• Characterize: protein inspection, search, alignment, profiles, embeddings, substitution scoring, stability-change prediction, secondary-structure and function-region prediction, protein completion, enzyme candidate retrieval, structure prediction, genomic context.
• Investigate: reactions and routes, thermodynamics, enzyme-panel evaluation of activity prediction, structure comparison and preparation, binding pockets, molecular dynamics, metadynamics, trajectory and replica analysis, relative binding free energy, QM/MM and reaction free energy.
• Engineer: candidate fitting and selection, variant proposal, construct design and assembly, coding-sequence recoding.
• Measure: measurement intake, fits, and the conversion and activity that a batch product target requires.
• Operate: definition of rate parameters from measured kinetics, reaction simulation, fitting, uncertainty propagation, comparison with measurements, output accounting and comparison of input costs.

alambiq capability list
alambiq capability show assay.fit

The first command prints every capability. The second prints the request fields, the result, the method and the limits of assay.fit.

Reports

alambiq run --report PATH writes a report: one HTML file that lists each step with its result. The first image shows the report of kinetics-from-measurements: a table of the readings and a Michaelis-Menten fit with the interval of the curve. The second image shows the search and alignment steps of the report of lanmodulin-comparison.

Installation

Install the package from PyPI (Python 3.11 or newer). The quantitative extra installs NumPy, SciPy and scikit-learn:

pipx install "alambiq[quantitative]"

The examples are in the repository. To run one from a clone, install the core in a virtual environment:

python -m venv .venv
.venv/bin/pip install -e ".[quantitative]"
.venv/bin/alambiq run examples/kinetics-from-measurements/campaign.yaml --report report.html

The example fits a Michaelis-Menten curve to a table of 10 initial rates and writes a report.

Each engine environment contains one of the following scientific packages and the Alambiq engines that call it: MMseqs2, P2Rank, ThermoMPNN, ESM, OpenMM, Boltz-2, OpenFF, OpenFE, pymbar, PySCF (QM/MM). Each engine environment except the Apple GPU environment of Boltz-2 is available as a Docker image, and each is available as a native environment built from a recipe. The installation guide describes both routes and the pinned resources. alambiq env check reports which engine environments are ready, and alambiq resource list reports which resources are installed.

Documentation

• Index: the pages by subject.
• Getting started: installation of the core and the first run.
• Installation: extras, engine environments, resources, run directories and platforms.
• Concepts: capabilities, families, campaigns, steps, bindings, presets, requests, results, records, runs, run records, reuse, engine environments, resources and reports.
• Python interface: campaigns and capabilities from Python.
• Examples: worked examples.
• Methods: choosing and reading the capabilities of each method area.
• Reports: the contents of a report.
• Licensing: the licenses of Alambiq and of optional software, models and datasets.

Repository layout

• src/alambiq/core: the campaign runner.
• src/alambiq/capabilities: the capabilities by family, with their scientific computation in functions that tests call directly.
• src/alambiq/engines: the engines that run inside engine environments.
• src/alambiq/report: the report writer.
• environments/: the recipes and verification files of the engine environments, and the Dockerfiles in environments/docker.
• examples/: the runnable examples.
• tests/core: the test suite (python -m pytest tests/core).

Models, engine environments and databases are installed as described in docs/install.md.