🔬 Cryo-Correlator: The 2D Class Detective 🕵️

Ever stared at hundreds of fuzzy gray squares from your cryo-EM run and wondered, "Are... are you two related?" Now you can find out!

Cryo-Correlator is a desktop tool with a simple GUI that dives into your 2D class average images and finds the hidden similarities between them. It's like a dating app for your protein views, but with more math and less awkward small talk.

(You can replace this with a screenshot of your actual GUI!)

🤔 What's the Big Idea?

In cryo-electron microscopy, we get thousands of images of individual particles, which are then averaged into "2D classes." These classes represent different views of a molecule. Sometimes, you get very similar-looking classes from different images or even within the same one.

Finding these similarities is crucial for:

• Validating your structure: Do different subsets of data produce the same views?
• Sorting through heterogeneity: Are there different conformations or populations of your particle?
• Cleaning up your dataset: Identifying and grouping the best, most consistent views.

Doing this by eye is slow, tedious, and subjective. This tool automates the process!

✨ Key Features

• Load Anything: Don't have just 3 images? Have 5? 10? Go crazy! Load as many grid images as you want.
• Dynamic Square Detection: No more rigid 10x10 grids. The app uses OpenCV to automatically find every single class average square in your images, no matter the layout.
• Ignores Text (Like a Boss): Those pesky resolution and particle count numbers written on the squares? The script automatically detects and masks them out, so it only compares the actual image data.
• Rotational Matching: Is your particle view slightly tilted in one square? No problem. The script rotates each image to find the best possible match, giving you a true similarity score.
• Interactive Heatmap: Visualize the entire 300x300 (or NxN) correlation matrix. The brighter the spot, the hotter the match! Click any pixel to instantly see the two squares being compared.
• Top Pairs Gallery: Instantly see a gallery of the most similar pairs, perfect for a quick overview of your data's consistency.

⚙️ How It Works: The Secret Sauce

The tool follows a sophisticated image processing pipeline to get you the most accurate results.

  \[Image 1\]      \[Image 2\]      \[Image N\]  
       |              |              |  
       V              V              V  

+---------------------------------------------------+
| 1. LOAD & DETECT SQUARES (OpenCV) |
| Finds all the little boxes in your big images. |
+---------------------------------------------------+
|
V
+---------------------------------------------------+
| 2. PROCESS EACH SQUARE INDIVIDUALLY |
| |
| [Square] -> [Mask Text] -> [Apply Circle Mask] |
| ^ | | |
| | V V |
| Original (Ignore this!) (Prep for rotation) |
+---------------------------------------------------+
|
V
+---------------------------------------------------+
| 3. ROTATIONAL CORRELATION (Sci-Image) |
| |
| FOR each pair of squares (A, B): |
| max_corr = -1 |
| FOR angle in 0..360: |
| B_rotated = rotate(B, angle) |
| corr = NCC(A, B_rotated) |
| IF corr > max_corr: max_corr = corr |
| |
+---------------------------------------------------+
|
V
+---------------------------------------------------+
| 4. VISUALIZE RESULTS |
| |
| [Interactive Heatmap] & [Top Pairs Gallery] |
| |
+---------------------------------------------------+

🚀 Getting Started

Ready to play detective with your own data? It's easy to get started.

1. Prerequisites

You'll need Python 3 and a few scientific libraries. You can install them all with pip:

pip install opencv-python scikit-image numpy matplotlib seaborn Pillow

2. Run the Script

Save the code as a Python file (e.g., cryo_correlator.py) and run it from your terminal:

python cryo_correlator.py

3. Using the App

1. Click "Load Grid Image(s)" and select one or more of your 2D class average files.
2. Wait for the status bar to confirm how many squares were found.
3. Click "Run Cross-Correlation". This may take a moment, especially with many squares, as it's performing thousands of comparisons!
4. Explore your results!
   • Click on the Correlation Matrix to inspect any specific pair.
   • Browse the Top Correlations tab for a quick look at the best matches.
   • Use the controls on the left to filter the matrix by a correlation threshold.

Happy correlating! May you find all the hidden gems in your datasets. ✨