KORA (Kinetic Ordered Regulatory Analysis)

Overview

KORA is a novel computational biology pipeline designed for inferring Gene Regulatory Networks (GRNs) from transcriptomic data, with a specific focus on neurodegenerative disorders. Leveraging principles from neuroscience and machine learning, KORA redefines GRN inference as a temporal causal learning problem. It utilizes Spiking Neural Networks (SNNs) trained with generalized Spike-Timing Dependent Plasticity (STDP) to capture dynamic, causal relationships between genes.

The pipeline spans data acquisition and preprocessing, spike encoding, SNN training, GRN extraction, and high-performance inference via CoreML on Apple's Neural Processing Unit (NPU). Comprehensive documentation, automated testing, and dynamic visualizations ensure reproducibility and interpretability.

Core Features

• Causal GRN Inference: Employs generalized STDP in SNNs to infer directed gene regulatory links based on temporal correlations.
• Disease-Specific Cohorts: Processes diverse transcriptomic datasets from public repositories (e.g., GEO) across various neurodegenerative diseases.
• Modular Pipeline: Structured Python scripts for data handling, SNN training, and GRN extraction.
• High-Performance Inference: Exported GRN operators as inference-only CoreML models for accelerated execution on Apple Silicon (NPU/CPU).
• Swift Integration: Native Swift application for loading CoreML models, performing batch inference, and benchmarking performance.
• Comprehensive Documentation: Detailed README.md files within each directory explain structure, manifest, and usage.
• Dynamic Visualizations: Tools for generating interactive GRN graphs, weight distribution plots, and performance benchmarks.

Pipeline Overview

1. Data Processing: Clean and normalize raw transcriptomic data.
2. Spike Encoding: Convert gene expression into spike trains.
3. SNN Training: Train cohort-specific SNNs using generalized STDP.
4. GRN Extraction: Extract directed GRNs from learned synaptic weights.
5. CoreML Export: Distill GRN operators into inference-only CoreML models.
6. Swift Inference: Perform high-performance inference on Apple devices (CPU/NPU).
7. Analysis & Visualization: Generate insights and dynamic plots.

Project Status

Version: v1.0.1 (Frozen for initial paper draft / release candidate)

All core pipeline components are implemented, tested, and documented. The system is capable of processing neurodegenerative cohorts, inferring GRNs, and performing high-speed inference.

Getting Started

1. Clone the Repository:

   git clone https://github.com/Sulkysubject37/kora.git
   cd kora

2. Setup Environment:

   python3 -m venv kora_env
   source kora_env/bin/activate
   pip install -r requirements.txt

3. Run Pipeline: Follow the instructions in the scripts/README.md to execute the data processing, encoding, training, and GRN extraction steps.
4. CoreML Export & Swift Inference: Refer to models/README.md and swift/README.md for details on exporting CoreML models and running the Swift inference engine.

Documentation

Each major directory (configs/, data/, docs/, models/, results/, scripts/, src/, swift/, tests/) contains a README.md file detailing its purpose, contents, and usage.

License

This project is licensed under the MIT License.

Citation

If you use this project in your research, please cite: MD.Arshad