biocompute 0.1.5

Wet lab automation as Python code. Maintained by london biocompute.

biocompute is a framework that lets you write wet lab experiments as plain Python. Define your protocol with calls like well.fill(), well.mix(), and well.image(). Then execute on real lab hardware that handles the liquid dispensing, mixing, and imaging automatically. No drag-and-drop GUIs, no manual pipetting.

If you know Python, you can run wet lab experiments.

Quick start

Create a virtual environment and install the biocompute package.

python -m venv .venv
source .venv/bin/activate
pip install biocompute

Create a file called super_simple_experiment.py and copy the code snippet.

from biocompute import wells, red_dye, green_dye, blue_dye

def experiment():
    for well in wells(count=3):
        well.fill(vol=80.0, reagent=red_dye)
        well.fill(vol=40.0, reagent=green_dye)
        well.fill(vol=20.0, reagent=blue_dye)
        well.mix()
        well.image()

Visualize the experiment in your terminal.

biocompute visualize super_simple_experiment.py

To run experiments on real hardware, log in and submit.

biocompute login
biocompute submit super_simple_experiment.py --follow

Results stream back to your terminal as experiments finish executing on the physical hardware.

Note: Hardware access is currently limited. Run biocompute login to join the waitlist.

How it works

Your experiment function describes intent. The compiler takes this high-level declarative code and turns it into a fully scheduled, hardware-specific protocol. It handles:

• Automatic parallelism — independent operations are identified and scheduled concurrently so protocols finish faster without any manual orchestration.
• Plate layout — wells are assigned to physical plates based on thermal constraints. Multi-temperature experiments get split across plates automatically.
• Operation collapsing — redundant per-well instructions (like 96 identical incubations) are collapsed into single plate-level commands.
• Device mapping — every operation is matched to the right piece of hardware (pipette, camera, incubator, gripper) based on a capability model, so swapping equipment never means rewriting your protocol.
• Multi-plate scaling — protocols that exceed a single plate are transparently distributed across as many plates as needed.

You describe what should happen. The compiler figures out how to make it fast.

Operations

wells(count=n) yields n wells. Multiple calls produce non-overlapping wells.

Reagents

Import the built-in reagents you need.

from biocompute import red_dye, green_dye, blue_dye, water

Because it's just Python

Use numpy. Use scipy. Use whatever. The system only sees wells and operations.

Colour sweep

Sweep red dye volume across ten wells using numpy to generate the range.

import numpy as np
from biocompute import wells, red_dye, green_dye, blue_dye

def experiment():
    for well, r in zip(wells(count=10), np.linspace(10, 100, 10)):
        well.fill(vol=r, reagent=red_dye)
        well.fill(vol=50.0, reagent=green_dye)
        well.fill(vol=50.0, reagent=blue_dye)
        well.mix()
        well.image()

Closed-loop optimisation

Submit an experiment, read results, use them to parameterise the next one.

import numpy as np
from scipy.interpolate import interp1d
from scipy.optimize import minimize_scalar
from biocompute import Client, wells, red_dye, green_dye

with Client() as client:
    volumes = np.linspace(10, 100, 8)

    def experiment_sweep():
        for well, v in zip(wells(count=8), volumes):
            well.fill(vol=v, reagent=red_dye)
            well.fill(vol=50.0, reagent=green_dye)
            well.mix()
            well.image()

    result = client.submit(experiment_sweep)

    model = interp1d(volumes, result.result_data["scores"], kind="cubic")
    optimum = minimize_scalar(model, bounds=(10, 100), method="bounded").x

    def experiment_refine():
        for well, v in zip(wells(count=5), np.linspace(optimum - 10, optimum + 10, 5)):
            well.fill(vol=v, reagent=red_dye)
            well.fill(vol=50.0, reagent=green_dye)
            well.mix()
            well.image()

    final = client.submit(experiment_refine)