How to Build a Model with AtomWorks: A Step-by-Step Guided Tutorial

Whether you’re training your first machine learning model or your tenth, preparing your data and building the infrastructure around your model can be a significant part of the work. For researchers developing machine learning tools for protein design, AtomWorks simplifies that process.

The new “How to Build a Model with AtomWorks” tutorial provides a hands-on introduction to the AtomWorks Python library and some of its most useful functionality for ML model development.

What is AtomWorks?

AtomWorks is a Python library that provides tools for working with biomolecular data and building the data pipelines needed for machine learning applications. It can help streamline tasks such as preparing and organizing your data or creating transform pipelines that convert your data into the format your model needs.

The tutorial provides a starting point for learning how these tools fit together in a model-development workflow.

What does the “How to Build a Model with AtomWorks” tutorial cover?

The tutorial takes you through the process of using AtomWorks to prepare data and train a machine learning model. You’ll learn how to:

• Clean and organize your data using AtomWorks
• Create a transform pipeline
• Build a dataset for your model
• Create and train a graph neural network (GNN) model using the resulting dataset

The GNN is used as an example to demonstrate how AtomWorks can fit into a complete machine learning workflow. However, the methods discussed can be applied to any machine learning model. The tutorial focuses on using AtomWorks rather than explaining the underlying model architecture, making it a useful starting point for researchers who want to learn how to incorporate the library into their own projects.

What if I’m New to AtomWorks?

If you’re new to AtomWorks, this tutorial is a great place to start. It introduces some of the library’s core functionality in the context of a practical machine learning application, giving you a foundation to build on as you explore the rest of the AtomWorks documentation.

How Do I Find the Tutorial?

We’ve created a short video introduction to the tutorial that provides an overview of what you’ll learn and how AtomWorks can help optimize the process of  preparing your data for model training.

To get started: Watch the video introduction

Ready to dive in?  Explore the full “How to Build a Model with AtomWorks” tutorial

The tutorial is a collaboration between the Rosetta Commons Tech Team and the Institute for Protein Design at the University of Washington.

Featured image inspired by PDB 11CE.