---
license: mit
task_categories:
  - tabular-classification
tags:
  - biology
  - boolean-networks
  - gene-regulation
  - systems-biology
  - experimental-design
  - panel-design
pretty_name: Accelerating Returns in Gene Panel Design
size_categories:
  - n<1K
---

# AcceleratingReturns: Data & Results

Checkpoints, results, and figures for: **Accelerating Returns in Gene Panel Design: A Completion Theory for Regulatory Network Identification**  
*BioSystems* (Elsevier), 2026.

Code repository: [https://github.com/Laddaphone/AcceleratingReturns](https://github.com/Laddaphone/AcceleratingReturns)

## Contents (73 files)

### Data

* `data/parsed_networks.pkl` — 285 parsed Boolean regulatory networks from [Biodivine Boolean Models](https://github.com/sybila/biodivine-boolean-models) (N=5–1076 genes). Each network is a dict with keys: `N`, `inputs`, `in_degrees`, `source_nodes`, `functions`, `nodes`, `filepath`.

### Results — Original Submission

* `results/original/` — 30 checkpoint files: theorem validation, strategy comparison, structural analysis, hitting sets, case study, ablation sweeps

### Results — Revision Experiments

* `results/revision/exp_protocol2.pkl` — Protocol 2 analysis (realistic observation budgets, n=171 networks)
* `results/revision/exp_dropout.pkl` — Dropout robustness (p ∈ {0.05–0.50}, n=50 networks)
* `results/revision/exp_scalability.pkl` — Computational scaling (N=50–2000, O(N³·¹⁵))
* `results/revision/exp_full_comparison.pkl` — Full 285-network strategy comparison
* `results/revision/exp_relaxed.pkl` — Error-tolerant (ε-relaxed) completion
* `results/revision/exp_theorem_gap.pkl` — Theorem prediction vs algorithm performance

### Figures

* `figures/paper/` — 7 paper figures (PDF, 600 DPI)
* `figures/revision/` — 14 revision figures (PDF + PNG)
* `figures/diagnostics/` — 15 diagnostic/validation PNGs

## Loading

```python
import pickle

with open("data/parsed_networks.pkl", "rb") as f:
    data = pickle.load(f)

networks = data["parsed"]  # list of 285 dicts
print(f"{len(networks)} networks, N={min(n['N'] for n in networks)}–{max(n['N'] for n in networks)}")

# Load a revision experiment
with open("results/revision/exp_dropout.pkl", "rb") as f:
    dropout_results = pickle.load(f)
```

## Citation

Douangnouanexay, L. (2026). Accelerating Returns in Gene Panel Design: A Completion Theory for Regulatory Network Identification. *BioSystems*.

## License

MIT