Research simulator for plant breeding with a Python/Qt GUI and a NumPy engine. It models repeated selection and crossing of plants using genomic estimated breeding values (GEBVs).
The genetic linkage engine is maintained in this repository alongside the simulator. Genetic linkage describes the tendency of nearby loci on a chromosome to be inherited together.
Developed as part of an undergraduate research project at Iowa State University. See the research poster.
With Python 3.9 or newer, install the engine from the repository root:
python -m pip install .
This installs genetic_linkage with NumPy only. The GUI is not needed to use it:
import numpy as np
from genetic_linkage import cross2
L1 = np.array([[0, 0, 0], [1, 1, 1]], dtype=np.int8)
L2 = np.array([[1, 1, 1], [0, 0, 0]], dtype=np.int8)
RF = np.array([0.05, 0.05])
progeny = cross2(L1, L2, RF, 3) # Shape: (3, 2, 3)To install the optional GUI dependencies and launch the existing simulator, run these commands from the source checkout:
python -m pip install ".[gui]"
python main.py
Calculations and data preparation run in worker threads. Qt handles status and
plot updates on the main thread and queues redraws. Pause and Stop take effect
between calculation steps. Closing requests cancellation and waits for the
worker to finish without blocking the event loop. File loading supports MATLAB
HDF5 .mat files.
Synthetic poster-scale dataset. The bundled demo data models 300 plants from 20 founders, with 14,000 loci in 100 linked blocks. The Python generator creates the NumPy inputs and runs a seeded reference simulation to illustrate the poster's score scale and selection curve. This is synthetic data, not the original research dataset.
The binary .mat file is stored with Git LFS. Install Git LFS and retrieve
the data before loading it:
git lfs install
git lfs pull
In the GUI, choose File > Open Data File, select
demo\synthetic-poster-demo.mat, then press Start.
The metadata records the generation parameters,
dataset seed 2017, simulation seed 43, and expected mean GEBVs.
The reference run rises from about 4.67 million to 8.04 million and levels off
in later generations. Exact reference results require np.random.seed(43)
before starting the simulation. Normal interactive runs can vary because the
GUI does not automatically apply the seed from the metadata file.
With the GUI dependencies installed, regenerate the bundled dataset and reference results from the repository root:
python scripts\generate_synthetic_poster_demo.py `
--output-path demo\synthetic-poster-demo.mat `
--metadata-path demo\synthetic-poster-demo.json `
--overwriteBoth paths are required and include the filename. They can use different
directories, and missing parent directories are created automatically.
The example uses --overwrite to replace the bundled files. Without it, the
generator refuses to replace existing output files.
The input arrays and reference scores are reproducible from the recorded seeds.
HDF5 container bytes can depend on the installed library versions.
from GenericLinkage import cross2 still works for existing callers, but new
code should use genetic_linkage. There is only one implementation of cross2.
Run the engine tests without GUI dependencies:
python -m unittest discover -s tests -v
With the GUI dependencies installed, run the offscreen UI regression tests:
python -m unittest discover -s tests\gui -v
These cover GUI-thread rendering, event-loop responsiveness, pause and restart, stopping and closing, background data loading, and dataset regeneration.
The bundled engine retains its original MIT license. That license notice covers the linkage code, not a new license grant for the rest of the simulator.
- Takao Shibamoto - main developer
- Dr. Lizhi Wang - mentor
- ISU Industrial and Manufacturing Systems Engineering
- ISU University Honors Program