An End-to-end Deep Learning Model for CyTOF data

Zicheng Hu, Ph.D.
Research Scientist
ImmPort Team
The Unversity of California, San Francisco

Introduction

This GitHub repository contains a tutorial for creating deep learning models tailored to CyTOF data. We will apply the model to diagnose latent cytomegalovirus (CMV) infection. We will also use a decision tree-based method to identify cell subsets that are associated with the CMV infection.

---

Install dependencies

Navigate into the tutorial folder. Create conda environment with dependencies for the tutorial. Install time takes around 5 mins.

conda create --name CyTOF_DL --file requirements.txt

---

Main tutorial

Activate the conda environment.

conda activate CyTOF_DL

Run the main tutorial DeepLearning_CyTOF.ipynb. It takes around 40 mins to run in a laptop with 2.5 GHz GPU (Intel Core i7) and 16 GB Memory.

jupyter notebook DeepLearning_CyTOF.ipynb

---

CyTOF data processing

Please see the script in FCS_to_Array folder for detailed preprocessing steps of the CyTOF data.

---

More information

For more background information of deep learning and its use in cytometry data, see the slides of the FOCIS 2019 workshop.

---