SDS-PAGE Analysis

This repository contains code and data for SDS-PAGE image analysis.
It includes functionality to:

• Detect band intensities
• Perform normalization
• Calculate 7s and 11s ratios
• Apply machine learning classification to predict the class of validation samples

---

📂 File and Folder Overview

• sds_images/
  Contains all SDS-PAGE images (.jpg or .jpeg).

• function.py
  Python script for:

  • Processing SDS-PAGE images
  • Detecting band intensities
  • Normalization
  • 7s/11s ratio identification
  • Machine learning pipeline

• bands_detection.ipynb
  Jupyter Notebook to run function.py.

• classification_file.csv
  Contains class labels for each sample.

• machine_learning_for_KNN_DT_RF_GB.ipynb
  Notebook with ML code for kNN, Decision Tree, Random Forest, and Gradient Boosting.

• Generated after running bands_detection.ipynb:

  • image_analysis/ – Each image has its own folder containing:
    • Detected 4 bands on marker lane
    • Vertical and horizontal lines
    • Minimum detected lines per image
  • Detected_bands_automatic/ – Automatically detected band images
  • manual_7s_11s_picture/ – Manually detected bands across all images
  • 7s_11s_peak_ranges.csv – 7s and 11s ratio across all images
  • mean_and_sum_of_7s_11s_ratio.csv – Mean and sum of 7s/11s ratios
  • normalized_df.csv – Normalized intensities of all samples + validation samples
  • normalized_samples.csv – Normalized intensities of samples only
  • normalized_validation.csv – Normalized intensities of validation samples only
  • ready_for_ml_class_1.csv – Final normalized dataset for ML
  • ready_for_normalization.csv – Data before quantile normalization

---

🔄 Workflow Overview

1. Run bands_detection.ipynb

   • Generates image_analysis/ and other processed outputs

2. Machine Learning

   • Normalized values are used to train models:
     • SVM
     • kNN
     • Random Forest
     • Decision Tree
     • Gradient Boosting
   • Models predict the classes of validation samples (unlabeled images)

3. 7s/11s Ratio

   • Ratios are automatically calculated from detected bands
   • Results saved in:
     • 7s_11s_peak_ranges.csv
     • mean_and_sum_of_7s_11s_ratio.csv

---

⚙️ Customization

Intervals

By default, the script uses 100 intervals. To change this, modify line 12 in function.py:

mln = <your_desired_intervals>

Base Directory

By default, outputs are created in the current working directory. To change:

Path.cwd() / "new_directory_name"

---

📊 Machine Learning Models

We use the following models for classification:

• Support Vector Machine (SVM)
• k-Nearest Neighbors (kNN)
• Decision Tree (DT)
• Random Forest (RF)
• Gradient Boosting (GB)

Validation set predictions are made using the trained models to classify unlabeled SDS-PAGE images.

---

📦 Requirements

• Python 3.x
• Jupyter Notebook