Serval

Overview

Serval is a fully original decoding pipeline for multiplexed spatial transcriptomics developed by Jenkin Tsui at the University of British Columbia and BC Cancer Research Institute.
Serval is a modular framework that allows users to decode MERFISH images using various decoding methods, including MERlin (Emanuel et al., 2020), Simple Nearest Neighbor, and Cosine-optimized decoding.

Features

• High‐accuracy barcode decoding via cosine‐optimized pixel matching
• Modular design: easily swap in/out different chromatic correction and filtering steps

Installation

1. Clone the repository

   git clone https://github.com/Roth-Lab/serval.git
   cd Serval-Decode

2. Create and activate your conda environment

   conda create --name serval-dev python=3.9
   conda activate serval-dev

3. Install dependencies

   pip install -r requirements.txt

Usage

To decode a dataset with default settings:

python /examples/standard_run.py \
  /examples/config.json

Datasets

4T1 cell culture dataset: https://doi.org/10.5281/zenodo.15678420, 4T1 tissue dataset: https://doi.org/10.5281/zenodo.15678603

Citation

If you use Serval Decode in your work, please cite:

Tsui, J. “Serval: A Cosine‐Optimized Pixel Decoder for Spatial Transcriptomics.” In preparation, 2025.

License

Serval Decode is licensed under the modified Apache License, Version 2.0; see the LICENSE file for details.

Copyright © 2025  
Jenkin Tsui. All rights reserved.

Trademarks

All other trademarks referenced herein are the property of their respective owners.