Hondius Hantavirus Outbreak Dashboard

The power of preparation, and a group text of epidemiologists

I have a lot of wonderful epidemiologist friends, and when we group text it can range anywhere from family updates to career discussions, and not uncommonly to public health current events. Things veered in the latter direction recently, which turned into a pretty fun project. Here’s how it went down:

4:30 pm - Chatting about the Hantavirus outbreak situation
5:07 pm - I toss out the idea of making a tracking dashboard where you can watch the ship move along a map and have outbreak-relevant information overlaid on the map
6:52 pm - First version of the dashboard live for review

You can check out the live version here! Now that first version had a lot fewer bells and whistles than it does now after a week of iterations, but it was pretty cool to be able to get at least a skeleton dashboard up so quickly on the site.

You can play to see the ship move along its course and highlight key stops and events.

Building Blocks: Connecting Components Makes Things Faster

How was I able to get things up and running in just a few hours? Well, mostly because I’ve been poking away at fun projects for quite a while, slowly building my skills. For incorporating the sequencing data on the dashboard, I’ve been working with Pathoplexus data for a while, so I already had code to pull in data through their API and create a few different charts. All I had to do was make some small updates to pull in Andes virus, create a list of outbreak sequences, and those pieces were ready to go! Also, I make liberal use of AI tools - for this project I was mostly working in VSCode using Codex.

Building Blocks: Embed rather than rebuild

The second thing I did to speed things up - don’t reinvent the wheel! There are already virology experts that have created public phylogenetic trees, so instead of trying to build and maintain my own, I reached out to Emma Hodcroft to get the thumbs up to embed her Nextstrain trees in the dashboard. They look beautiful! Andes virus is is a segmented viruse with three segments (S, M, L) and I was able to embed all three segment trees as three tabs, so it’s easy to just click between the three trees. I love embedding trees in pages because it lets you add commentary and links, making it much easier to guide people toward good interpretation of the data.

We gratefully acknowledge Emma Hodcroft, EVE Group: Swiss TPH, University of Basel, Neherlab: Biozentrum, University of Basel and the Pathoplexus Team for creating and maintaining the Nextstrain builds embedded here, as well as nextstrain.org.

What do the sequences tell us?

So, with all that data on there - what does it actually tell us? Well, thankfully it’s overall pretty boring! There are very few differences between all the sequences, which aligns with the epidemiological data which suggests that there was a single exposure and then transmission between people on the ship. If the viruses were more different it could indicate that there were multiple exposures - nobody really expected to see that, so it was nice to see the sequences align with the epidemiological hypothesis. Now, the genomic data can’t tell you for example if case 2 was exposed directly to the original zoonotic source the same time as case 1, or transmission occurred between case 1 and case 2.

Current evidence suggests subsequent human-to-human transmission onboard the ship. This is also supported by a preliminary analysis of the sequences, which show a close, near-identical sequenced from different cases.

^{- WHO (https://www.who.int/emergencies/disease-outbreak-news/item/2026-DON601)}

Take home messages

I think this graphic is one of the most important, and hopefully it ages well. As of May 17th, all of the confirmed and probably cases are among people who were on the ship. These would all be considered primary cases in the outbreak. If any of their contacts (for example there are some flight contacts being monitored) become ill, those would be considered secondary cases. If the outbreak can be stopped with no secondary cases, that would be a great outcome. The incubation period is quite long (4-42 days), so monitoring will be ongoing for quite a while for ship passengers as well as any exposed contacts.

Wrapping it up

It’s been a fun project to connect with some epi friends, and keep working on my dashboarding and automation skills. I’ve learned about some new data sources - I switched over to using a linelist maintained by global.health instead of continuing to curate my own, and improved my automated connections with Pathoplexus sequencing data. I’m a big fan of leaning on the work of others when feasible! Focus on what you’re good at, and lean on the expertise of others where you can.
I’ll continue to keep this one updated over time, so keep visiting https://www.genomicepi.com/outbreaks/hantavirus-hondius/ to see future updates!

Thank you Cristin Young for your help reviewing and updating the initial linelist and ongoing suggestions for the dashboard!