Gene Visualization Tool (cellColor)

Can't wait for action? Watch the demo: https://youtu.be/FGlrxKn701Y

Download here: https://pypi.org/project/cellColor/

An interactive desktop application for visualizing spatial transcriptomics data, cell segmentation masks, cell clustering data and microscopy images. Perfect for verifying cell segmentation accuracy and exploring spatial gene expression patterns.

---

👥 Credits & Project Team

• Developer: Anthea Guo
• Mentor: Kushal Nimkar
• Principal Investigator (PI): Prof. Karthik Shekhar

---

Motivation

The motivation behind this was to simplify the workflow for new datasets by creating a flexible and fast tool that allows users to easily visualize spatial patterns of transcriptomic/cell segmentation/cell center data on histological sections.

---

✨ Features

• Image Loading & Zooming: Load tissue/microscopy images, zoom into regions, and reset to full view.
• Cellpose Segmentation Overlay: Overlay Cellpose-generated segmentation masks or outlines with smooth, cached zooming. This app allows you to automatically generate cellpose outlines from segmentation masks however this will take roughly 30 minutes to compute for the first time. Once the outline file exists locally, the visualization process will be much faster.
• Transcript Visualization: Import transcript coordinates (x, y, gene), align with images using transformation matrices, and overlay selected genes.
• Single-Cell Integration: Load AnnData cell center positions (.h5ad), toggle display, and customize appearance.
• Newest feature! Visualize cell types!: We have an inbuilt function that takes cell type labels and matches it with the cellpose segmentation masks. Users can select cell types and the corresponding masks will appear.
• User-Friendly Toolbar: Intuitive controls for overlays and zoom, live status feedback, and collapsible navigation frames.
• Other features: Load transformation matrices for accurate transcript-image alignment. Program is designed high speed performance which faster file versions saved after first run, including parquet files for transcripts, downsized tiff files and saved outline files. Zoom is also cached so that the user can go back to the previous zoom option.

---

🚀 Installation

Option 1: Install via PyPI (v0.1.0) When using a MacOS system that is more recent only pip3 is available, if this doesn't work switch back to pip.

pip3 install cellColor

Release: Nov 20, 2025 (PyPI link)

Launch:

cellColor

Option 2: Local Development (Editable Mode)

git clone https://github.com/crocodile27/cellColor.git
cd cellColor
conda create -n cellcolor python=3.10
conda activate cellcolor
pip install -e .

Run locally:

cellColor

---

‼️ IMPORTANT: Required Data Formats (Email antheaguo@berkeley or kushalnimkar@berkeley with any questions)

Images

Tissue Section Images: .png, .jpg, .tif. High resolution tissues that will act as the canvas and which regions you choose is up to your discretion. Click on file-> load image and choose desired image. A downsized version of the image will be produced after it is loaded for the first time. Feel free to choose either of them in future runs but downsized version will be automatically chosen.

All other data:

Place all of the following files in a single folder with the format **[prefix]***rn[insert run number]*_**rg[insert region number]**. E.g. 140g_rn3_rg0 -> _this is really important for the auto_load_file function to correctly detect which run and region you are working on and automatically load all necessary files.

Files automatically loaded (upon clicking the folder and choosing open) include:

• Cellpose Masks: .npy image mask generated from cellpose. Outlines are optional
• Detected Transcripts: CSV/TSV/Parquet with barcode_id, global_x, global_z, x, y, fov, gene, transcript_id as columns. global_x & global_y will be used as the coordinates.
• Transformation Matrix: CSV/TSV for alignment of gene and tissue data.
• AnnData:.h5ad with cell center coordinates saved in global_x and global_y observations.

---

🧪 Example Workflow

1. Open the app: cellColor
2. Load image: File → Load Image to display tissue section.
3. Auto load all files: → click on desired folder
4. Overlay Data: choose the data you'd like to overlay including: cell centers, gene transcripts, cellpose masks/outlines, cell clusters.
5. Zoom & reset: Zoom into areas of interest; use Reset Zoom to return to previous zoom level.

For bugs & other issues email: antheaguo@berkeley.edu and kushalnimkar@berkeley.edu For sample data contact: kushalnimkar@berkeley.edu