Skip to content

RELEASE_NOTES.md

Latest commit

 

History

History
179 lines (127 loc) · 4.68 KB

RELEASE_NOTES.md

File metadata and controls

179 lines (127 loc) · 4.68 KB
Raw

Autism Pathway Framework v0.1.0

RESEARCH USE ONLY — Not for clinical decision-making

The first reproducibility release of the Autism Pathway Framework.

DOI

Quickstart

# Clone and setup
git clone https://github.com/topmist-admin/autism-pathway-framework.git
cd autism-pathway-framework
python3 -m venv autismenv && source autismenv/bin/activate
pip install -r requirements.lock && pip install -e .

# Run demo
make demo

Or try in Google Colab (no installation):

Open In Colab

What's Included

Core Framework (12 Modules)

Module Description
01 Data Loaders VCF, pathway, expression, constraint data
02 Variant Processing QC filtering, annotation, gene burden
03 Knowledge Graph Biological relationship encoding
04 Graph Embeddings TransE, RotatE embeddings
05 Pretrained Embeddings Geneformer, ESM-2, PubMedBERT
06 Ontology GNN Ontology-aware graph neural network
07 Pathway Scoring Multi-evidence pathway disruption
08 Subtype Clustering GMM clustering with validation
09 Symbolic Rules Biological rule engine (R1-R7)
10 Neurosymbolic GNN + symbolic integration
11 Therapeutic Hypotheses Drug-pathway mapping
12 Causal Inference SCM, do-calculus, counterfactuals

Integration Pipelines

  • Subtype Discovery: VCF → pathway scores → molecular subtypes
  • Therapeutic Hypothesis: subtypes → rules → drug candidates
  • Causal Analysis: individual mechanistic reasoning

Demo Runtime

Environment Runtime Notes
Local (M1 Mac) ~30 seconds Synthetic 50-sample dataset
Google Colab ~2 minutes Including setup
Docker ~45 seconds Containerized execution

Outputs

The demo pipeline generates:

outputs/demo_run/
├── pathway_scores.csv      # Pathway disruption scores per sample
├── subtype_assignments.csv # Cluster assignments with confidence
├── report.json             # Machine-readable summary
├── report.md               # Human-readable report
├── run_metadata.yaml       # Reproducibility metadata
├── pipeline.log            # Execution log
└── figures/
    └── summary.png         # Visualization (optional)

See docs/outputs_dictionary.md for interpretation guide.

Validation Gates

Every run includes automatic validation:

Gate Description Threshold
Negative Control 1 Label shuffle test ARI < 0.15
Negative Control 2 Random gene sets ARI < 0.15
Stability Test Bootstrap resampling ARI ≥ 0.8

Results appear in report.json and report.md:

{
  "validation_gates": {
    "all_passed": true,
    "tests": [...]
  }
}

Reproducibility

This release ensures deterministic execution:

  • Pinned dependencies: requirements.lock
  • Seed control: PYTHONHASHSEED=42
  • Golden outputs: tests/golden/expected_outputs.yaml
  • CI verification: GitHub Actions runs on every push

To verify reproducibility:

make reproducibility-test

Requirements

  • Python 3.10+ (3.11 recommended)
  • 16 GB RAM recommended
  • 5 GB disk space

Citation

@software{chauhan_2026_autism_pathway,
  author       = {Chauhan, Rohit},
  title        = {Autism Pathway Framework},
  version      = {0.1.0},
  year         = {2026},
  publisher    = {Zenodo},
  doi          = {10.5281/zenodo.18403844},
  url          = {https://doi.org/10.5281/zenodo.18403844}
}

Known Limitations

  • Demo dataset is synthetic (N=50)
  • GMM clustering assumes Gaussian mixture structure
  • Network propagation depends on PPI network quality
  • Not validated for clinical use

See CHANGELOG.md for full details.

Documentation

Links

RESEARCH USE ONLY — This framework is for hypothesis generation. Not for clinical diagnosis or treatment decisions.