Scalable Data Analysis with Dask

Lyda Hill Department of Bioinformatics

University of Texas Southwestern Medical Center

---

Course Information

Dates: 8/6/2025 Instructors: Kevin Dean, Ph.D. and Conor McFadden
Credit Hours: 0.5 Room: G9.102
Class Size: 10

---

Course Description

This one-day advanced Python course introduces graduate students and postdoctoral scientists to parallel and distributed computing using Dask, with a focus on large-scale biomedical data analysis. Participants will learn how to process and analyze 2D/3D image data and sequencing datasets using Dask Arrays and DataFrames, including out-of-core computation and storage with Zarr. The course covers local parallelism, distributed computing on SLURM clusters, and best practices for performance profiling and optimization. Through lectures, hands-on exercises, and real-world examples, attendees will gain the skills to scale their existing NumPy, Pandas, and scikit-image workflows to handle large datasets efficiently and reproducibly.

---

Course Objectives

By the end of this course, participants will be able to:

• Understand the need for parallel and distributed computing in handling large biomedical datasets.
• Learn what Dask is and how it enables parallel processing by dividing work into tasks.
• Familiarize with Dask’s core concepts (task graphs, schedulers, lazy evaluation) and how it integrates with Python’s scientific stack.

---

Installation of Dependencies

Using venv

python -m venv .venv
pip install --upgrade pip
pip install -e .

Using Conda on BioHPC

Clone www.github.com/TheDeanLab/dask-nanocourse to your BioHPC home directory, then run the following commands:

module load python/latest-3.12.x-anaconda
(base) module load gcc/8.3.0
(base) conda create -n dask-nanocourse python=3.10
(base) conda activate dask-nanocourse
(dask-nanocourse) pip install --upgrade pip setuptools wheel
(dask-nanocourse) conda install -c conda-forge pyzmq
(dask-nanocourse) pip install -e . --no-cache-dir

Add the kernel to your iPython environment and launch a Jupyter Notebook

Install the IPython kernel inside the activated virtual environment, then register it with Jupyter:

Using venv

python -m ipykernel install --user --name .venv --display-name "Dask Nanocourse"

Using Conda

python -m ipykernel install --user --name dask-nanocourse --display-name "Dask Nanocourse"

Install Jupyter Lab Extensions

conda install nodejs
jupyter labextension install dask-labextension

Launch Jupyter Lab

jupyter-lab