Legacy Components¶
Legacy code is kept under legacy/ for reproducibility. New development should not import legacy modules unless the work is explicitly about reproducing historical results.
Legacy Neural Ancestry¶
Path: legacy/neural_ancestry_predictor_deprecated/
There is no supported genomics neural CLI entrypoint. The code and configs remain in legacy/ only for historical inspection and reproducibility work.
For new ancestry/model experiments, prefer:
genomics genotype ...genomics variant ...genomics snp-ancestry ...
Implementation Structure¶
| Module | Purpose |
|---|---|
neural_ancestry_predictor_deprecated.py |
Historical monolithic script preserved in legacy code |
config.py |
Legacy YAML parsing and dataset path resolution |
data_pipeline.py |
Legacy processed-cache creation and loading |
models.py |
Historical NN/CNN model definitions |
training.py |
Training loop and checkpoint handling |
evaluation.py |
Test/evaluation helpers |
sklearn_baselines.py |
SVM/RF/XGBoost baseline support |
tune_sklearn_baselines.py |
Hyperparameter search for sklearn baselines |
interpretability.py, annotate_deeplift_windows.py |
DeepLIFT/interpretability utilities |
verify_processed_dataset.py |
Dataset consistency checks |
The legacy package now imports shared infrastructure from genomics.core where possible, but the model/data assumptions remain historical.
Why It Remains¶
This code produced historical benchmarks and plots. It remains available so old results can be reproduced or compared, but it is not the recommended implementation for new experiments.
Legacy-only configs live beside this code under legacy/neural_ancestry_predictor_deprecated/configs/. Configs for reproducing the old raw-center-crop representation on the active genotype predictor live under configs/predictors/genotype_based/neural_legacy/.
Legacy Longevity Dataset¶
Path: legacy/neural_longevity_dataset/
This code remains coupled to historical top3 layout assumptions and is not the recommended path for new dataset builds.
Implementation Structure¶
| Module | Purpose |
|---|---|
neural_longevity_dataset.py |
Historical dataset construction pipeline |
longevity_train.py |
Historical training script for longevity experiments |
Use the active dataset builder under src/genomics/workflows/dataset_builders/non_longevous/ for new dataset creation.