What is it about?

Understanding exactly how water freezes is a major scientific challenge, especially when it's squeezed into tiny spaces like nanotubes, where it can form exotic types of ice. While computer simulations can model this process, they only give us a huge list of atom positions. Figuring out what kind of ice has actually formed was often a slow, manual process of "by eye" inspection. To solve this, we developed and implemented for the first time a series of novel algorithms based on graph theory. Instead of just looking at the positions of atoms, our method analyzes the topological network of hydrogen bonds—the web of connections between water molecules. This allows us to precisely identify the complex building blocks of ice, like "hexagonal cages" and "double-diamond cages." We created d-SEAMS, a free and open-source software tool that acts like an automated "digital microscope" for this simulation data. It uses powerful algorithms to automatically detect and classify the many different forms of ice, from the familiar hexagonal structure to strange new phases that appear only in nanoconfinement. Crucially, d-SEAMS is built using a system called nix, which completely solves the frustrating problem of software installation, guaranteeing that it runs perfectly and reproducibly on any computer, from a laptop to a supercomputing cluster.

Featured Image

Why is it important?

Implementing state of the art algorithms for structural analysis based on connectivity motifs, d-SEAMS allows researchers to study complex freezing events that were previously too difficult or time-consuming to analyze. It turns raw simulation data into real physical insights about how ice forms. It replaces subjective visual inspection with objective, quantitative metrics, making the science more rigorous and reliable. The use of nix also solves the "dependency hell" that plagues scientific software, ensuring that an analysis performed today can be perfectly reproduced by another scientist years from now. d-SEAMS saves researchers immense amounts of time and effort, allowing them to focus on scientific discovery rather than wrestling with software.

Perspectives

This project was my first work as a co-lead author, with the brilliant Amrita Goswami, who pioneered the algorithms demonstrated in separate publications and implemented them elegantly too. It was where I first truly combined my passion for fundamental scientific questions with robust software design. I wanted to build a tool that solved two problems at once: the scientific challenge of automatically identifying complex ice structures, and the practical software challenge of making it work reliably for everyone, everywhere. It's been half a decade since d-SEAMS was first created, and it has been incredibly rewarding to see the project thrive, receiving grants from programs like Google Summer of Code and evolving over time. What I'm most proud of is that the core algorithms and the original design philosophy have remained solid. For me, d-SEAMS embodies the idea that solving the "boring" problems of software engineering—like reproducibility and dependency management—is not separate from doing good science. It's essential to it, and this project set the foundation for how I've approached my work ever since.

Rohit Goswami
University of Iceland

Read the Original

This page is a summary of: d-SEAMS: Deferred Structural Elucidation Analysis for Molecular Simulations, Journal of Chemical Information and Modeling, March 2020, American Chemical Society (ACS),
DOI: 10.1021/acs.jcim.0c00031.
You can read the full text:

Read

Resources

Contributors

The following have contributed to this page