Image Credit: Susanna Hamilton
Scientific Frontline: Extended "At a Glance" Summary: Delphy Software for Viral Variant Tracking
The Core Concept: Delphy is a web-based software application designed to rapidly assemble phylogenetic trees, mapping the evolution and spread of viral variants from genomic sequence data in near-real time.
Key Distinction/Mechanism: Delphy operates entirely within a web browser, processing data locally on the user's device without requiring specialized infrastructure or an internet connection after loading. It achieves analysis speeds 100 to 1,000 times faster than existing methods by streamlining and optimizing the underlying statistical models, completing tasks that previously took months in hours.
Origin/History: The software was conceptualized mid-2020 by Patrick Varilly, Pardis Sabeti, and Ben Fry to address the computational bottleneck of analyzing vast quantities of SARS-CoV-2 genomic data during the COVID-19 pandemic.
Major Frameworks/Components:
- Bayesian phylogenetics
- Interactive data visualization algorithms
- Web browser-based local execution environments
Branch of Science: Computational Biology, Epidemiology, Genomics, Virology
Future Application: Delphy is intended for widespread deployment among frontline public health officials and scientists globally, enabling them to independently conduct advanced outbreak tracking and lineage identification to inform rapid containment strategies for diseases such as Ebola, Zika, mpox, and H5N1.
Why It Matters: As large-scale viral sequencing becomes standard during outbreaks, traditional computational tools are too slow to inform immediate public health interventions. Delphy democratizes complex phylogenetic analysis, translating raw genetic data into actionable insights fast enough to potentially halt the spread of emerging pathogens.
It was mid-2020, and Patrick Varilly, a software engineer and data scientist, was stuck at home, eager to help the world navigate the ongoing COVID-19 pandemic. He reconnected with Pardis Sabeti, a core institute member of the Broad Institute who was at the forefront of analyzing how the SARS-CoV-2 virus was spreading, and with Ben Fry, her longstanding collaborator and principal at Fathom Information Design, a software firm known for tackling complex data problems. Varilly had worked closely with Sabeti and Fry at MIT more than 20 years earlier.
At the time, Sabeti, Fry, and their teams were studying thousands of SARS-CoV-2 genomes from COVID-19 patients to reconstruct the path of viral transmission and identify which viral variants were emerging. Normally, retracing that path—by mapping how different variants are genetically related to each other in what is called a phylogenetic tree—takes a significant amount of time and computing power.
Varilly, Sabeti, and Fry saw an opportunity to accelerate the process while making data accessible and easier to interpret. The result is Delphy, a new platform for rapid, interactive phylogenetic analysis. In a paper published in Nature, the researchers report how they rebuilt state-of-the-art phylogenetic tree models to make them faster, more efficient, and scalable while maintaining the models’ accuracy. Because Delphy runs entirely within a web browser, anyone with a laptop can perform these analyses without specialized training, software, or computing infrastructure.
The researchers demonstrated that Delphy could analyze 100,000 viral genomic sequences and create phylogenetic trees 100 to 1,000 times faster than existing methods, reducing a months-long task to one that takes hours. To use the platform, users input sequence data, click “run,” and then explore the resulting tree by moving across lineages, mutations, and time to see patterns of viral evolution. This makes complex outbreak analysis more accessible to public health teams working in real time.
“Doing large-scale sequencing of viruses is now the norm, but the old tools don’t work as well with such large numbers of sequences to analyze,” said Varilly, first author of the new paper and a computational research scientist in the Sabeti lab. “Public health officials have a lot to juggle, so the goal of Delphy is to simplify and speed up the process as much as possible while maintaining accuracy. This project was driven by the idea that genomic sequencing can be used earlier and help contain outbreaks faster.”
“The challenge was not just displaying an enormous amount of data but making it useful,” said Ben Fry, principal at Fathom Information Design. “Delphy gives people a way to move through the tree, ask questions of it, and learn about the tree by interacting with it.”
“To respond to outbreaks quickly, we need to empower frontline scientists and public health teams everywhere to carry out the most advanced analyses themselves,” said Sabeti, senior author of the study. “We are tremendously excited that Delphy can bring this capability directly to the front lines, allowing people anywhere to generate rigorous insights quickly and use them to guide faster, more effective outbreak response.”
Viral Family Tree
Viruses are constantly evolving and mutating. Often, the more a virus circulates, the more likely it is to acquire new mutations that can impact its ability to spread and infect people. Analyzing viral genomes to build phylogenetic trees allows scientists and public health officials to reconstruct outbreak timelines, visualize how and when one strain mutated into the next, and estimate how quickly the virus is changing and spreading. These insights can inform public health interventions; the faster experts can make these decisions based on accurate and reliable data, the more likely they are to contain outbreaks quickly.
Creating phylogenetic trees from thousands of viral genomic sequences using previous methods can take weeks or even months. Varilly and his team simplified and streamlined Delphy’s statistical calculations to speed up the analysis.
To test Delphy’s accuracy, the team reproduced analyses from recent outbreaks and epidemics, including Ebola, Zika, SARS-CoV-2, mpox, and H5N1. They report an increase of two to three orders of magnitude in the speed of analysis, showing that Delphy could create a phylogenetic tree from a dataset of 100,000 viral genomic sequences within a day. Delphy can also identify key viral lineages and mutations, as well as the timing of their emergence, with the same accuracy as other commonly used methods.
Importantly, Delphy runs all calculations locally, meaning sequence data never leaves the user’s computer. Once Delphy.bio loads, users can perform analyses without an internet connection.
Varilly says he hopes that the public health community will adopt Delphy to help with outbreak response. “It’s exciting to be at the forefront of developing new solutions that hopefully are not far away from being utilized everywhere,” he said.
Research material: Delphy.bio is now freely available to anyone who needs to map virus sequence data to understand the course of an outbreak.
Funding: This research was supported by Flu Lab; TED’s Audacious Project; the John D. and Catherine T. MacArthur Foundation; the US CDC Office of Advanced Molecular Detection (contract #75D30122C14365 and grant #NU50CK000629); National Institutes of Health and National Institutes of Allergy and Infectious Diseases (grants #U19AI110818, #U01AI151812, and #U19AI135995-07S2).
Published in journal: Nature
Title: Scalable near-real-time Bayesian phylogenetics for outbreaks with Delphy
Authors: Patrick Varilly, Mark Schifferli, Katherine Yang, Paul Cronan, Ivan Specht, Tim Burcham, Olivia Glennon, Olivia Jacks, Ellory Laning, Libby Marrs, Kyle Oba, Shannon Yeung, Karlie Wenran Zhao, Edyth Parker, Ifeanyi Omah, Jonathan E. Pekar, Laura Luebbert, Kristian G. Andersen, Daniel J. Park, Stephen F. Schaffner, Bronwyn L. MacInnis, Christian Happi, Jacob E. Lemieux, Al Ozonoff, Michael Mitzenmacher, Ben Fry, and Pardis C. Sabeti
Source/Credit: Broad Institute | Jessica Colarossi
Edited by: Scientific Frontline
Reference Number: cobi091626_01
