/btechproj

Primary LanguageJupyter NotebookMIT LicenseMIT

Classification of Highly Interacting Regions of the Genome with Explainable AI

Overview

This project aims to identify regions governing gene regulation and classify them as highly interacting or non-interacting using Explainable AI. By examining these sequences, we can classify genomic regions and determine influential genomic properties in the transformation of a cell into a diseased state.

Objectives

  • Model biological sequencing and DNA data to understand their underlying properties and patterns.
  • Use deep learning algorithms to distinguish highly interacting regions and their boundaries.
  • Classify genomic sequences as potentially interacting or non-interacting.
  • Identify traits of highly interacting regions that make them more interactive.
  • Understand how these specific regions influence gene regulation.
  • Find genomic sequence properties that influence a cell's transformation into a diseased cell.

Tech Stack

  • Programming Languages: Python
  • Libraries and Frameworks: TensorFlow, Keras, Sklearn, Matplotlib, Seaborn
  • Tools: Jupyter Notebook, Git
  • Data: Sub-kb Hi-C in D. melanogaster, ChIP-seq data from ENCODE project

Methodology

Phase 1 - Markov Models

Architecture

  1. Preprocessing Flow:

    • Generate dummy data with a probability of 0.25 for each nucleotide.
    • Create files embedding highly interacting regions using random functions.

  2. Training and Testing:

    • Train highly interacting region Markov models on respective files.
    • Perform cross-validation and visualize results with AUC ROC curves, accuracy, and F1 scores.

Phase 2 - Deep Learning

Deep Learning Model

  • Model Architecture:
    • Conv1D layers with Batch Normalization and LeakyReLU activation
    • MaxPooling1D and Dropout layers to prevent overfitting
    • Dense layers for final classification

Training and Testing

  • Perform cross-validation on simulated and real datasets.
  • Evaluate the model using metrics such as accuracy, AUC, and ROC curves.

Results

Markov Models

  • Simulated Data:
    • Generated dummy data with embedded highly interacting regions.
    • Cross-validation results showing accuracy and AUC scores.

Deep Learning

  • Simulated Data:
    • Achieved high testing accuracy with consistent results across folds.
  • Drosophila Data:
    • Moderate accuracy indicating room for improvement.

Directory Structure

  • Data/: Contains all datasets and related files.

    • fruitfly/: Data specific to fruit flies.
      • Fruitfly Bed files/: BED files for different shifts.
      • Fruitfly Datasets/: Dataset files for different shifts.
      • Fruitfly Fasta/: FASTA files for different shifts.
    • deepbind-exe-file/: Contains input files for DeepBind executions.
    • dummy_markov_data/: Contains Markov model data.

  • colab_files/: Contains Jupyter notebooks for various data processing and analysis tasks.

  • __MACOSX/: Contains system files for macOS, which are not needed for the project execution.

Setup and Installation

  1. Clone the repository:

    git clone <repository-url>

  2. Navigate to the project directory:

    cd btechproj-main

  3. Install required dependencies:

    • Ensure you have Python installed.
    • Install dependencies using pip: 
      pip install -r requirements.txt
    • Alternatively, if a requirements.txt file is not provided, manually install dependencies mentioned in the notebooks and scripts.

Data Description

  • Fruitfly Data: This includes BED, FASTA, and dataset files for different shifts (e.g., shift_200, shift_500).

    • BED files: Contains genomic regions data.
    • FASTA files: Contains sequences of DNA.
    • Dataset files: Contains various datasets used for analysis.

  • DeepBind Data: Input files for running DeepBind, a tool for predicting protein binding.

  • Markov Model Data: Files related to Markov models used in the analysis.

Jupyter Notebooks

Located in the colab_files/ directory, these notebooks provide various analyses and processing steps, such as:

  • Markov models with cross-validation
  • TensorFlow and PyTorch implementations
  • Binding site predictions
  • Pipelines for converting BED to FASTA and other tasks

Usage

  1. Running Jupyter Notebooks:

    • Navigate to the colab_files/ directory: 
      cd colab_files
    • Start Jupyter Notebook: 
      jupyter notebook
    • Open the desired notebook and follow the instructions within.

  2. Data Processing Scripts:

    • Data processing scripts are located within the Data/ directory. Run these scripts as needed for your analysis.

Contributing

  1. Fork the repository.
  2. Create a new branch (git checkout -b feature-branch).
  3. Commit your changes (git commit -am 'Add new feature').
  4. Push to the branch (git push origin feature-branch).
  5. Create a new Pull Request.

License

This project is licensed under the MIT License. See the LICENSE file for details.

Contact

For any inquiries or issues, please contact [Your Name] at [your-email@example.com].