We will keep a close look out, thanks to our academic and public health partnership that makes this possible. Here is a team selfie, including our project lead @drelyse.bsky.social and some of our awesome @sabetilab.bsky.social @broadinstitute.org & the Mass DPH team. (10/10)

By establishing and validating these methods, we were positioned to partner with @MassDPH and @MassDeptAgr to support mandatory H5N1 testing in milk from cattle dairy farms across the state. Starting from August 2023, and continuing today, regular testing has yet to reveal a positive result. (9/10)

At higher sample concentrations (>500 cp/uL), hybrid-selected RNA-Seq libraries produced the most complete genomes, while an Amp-Seq protocol performed best at lower concentrations. (8/10)

We were also able to produce near-complete genomes from even ultra-pasteurized milk using 3 library construction methods — RNA-Seq, hybrid-selected RNA-Seq, and Amp-Seq — and in samples with copy numbers as low as 14 H5N1 cp/uL. (7/10)

Of our 214 commercial milk samples, 55 (26%) tested positive for H5N1 RNA by dPCR, while 48 (22%) tested positive by qPCR. Most positive samples were from processing plants in four states with reported H5N1 outbreaks: Colorado, Idaho, Michigan, and Texas. (6/10)

We also assessed a variety of pre-processing and extraction protocols for H5N1 detection and sequencing. All three tested extraction kits demonstrated similar recovery, and fat content and pre-centrifugation did not significantly affect detection. (5/10)

With 214 store-bought milk samples from 20 states, we evaluated dPCR & qPCR assays for H5N1 w/primers targeting the H5 subtype of HA gene. All assays performed well w/minimal optimization – dPCR was slightly more sensitive – & bovine Ribonuclease P gene served as an effective positive control.(4/10)

The best time to set up testing is indeed before any cases, but setting up testing requires positive cases for validation. A classic conundrum, that here was solved by ready access to positive samples are our store shelves – commercial milk – and spurred by a collaboration with @BostonGlobe. (3/10)

With an H5N1 outbreak spreading through dairy cattle in multiple states, we looked to develop and optimize diagnostic and surveillance capabilities in Massachusetts before a single case had reached New England. (2/10)

More thank you's to: @kgandersen.bsky.social, @dannyjpark.bsky.social , @glipsnort.bsky.social , Bronwyn MacInnis, Christian Happy, Jacob E. Lemieux, Al Ozonoff, Michael D. Mitzenmacher, and @benfry.com 16/17

Thanks to everyone involved in this project: Patrick Varilly, Mark Schifferli, Katherine Yang, @timbrrr.bsky.social , Paul Cronan, Olivia Glennon, Olivia Jacks, Ellory Laning, Libby Marrs, Kyle Oba, Shannon Yeung, @edythparker.bsky.social , Ifeanyi Omah, Jonathan Pekar, @lauraluebbert.com 16/17

Now more than ever, we're excited to see how making Bayesian phylogenetics broadly usable for a much wider audience can enhance real-time tracking and strengthen public health responses to emerging outbreaks. 15/17

While Delphy focuses on early outbreak response, its core concept—EMATs—has broader potential. Any Bayesian phylogenetics calculation can be adapted to EMATs. We welcome community efforts to integrate EMATs into existing tools for everyone's benefit.
#Phylogenetics 14/17

Delphy rethinks Bayesian phylogenetics using EMATs instead of implicit trees + Felsenstein tree likelihood, opening the door to:
⚡ Faster outbreak analysis
🎯 Better sampling
🧩 Richer modeling
📊 New analysis tools
And we’re just getting started.
13/17