At the Nucleic Acid Observatory, we work on better detecting new pandemics. By now, I believe we can build something akin to a large-scale weather radar system, but for outbreak detection. Advances in AI make such a system both necessary and feasible. Here I lay out why:

RT @buxwal: Since October 2024 I've been working with SEVIS — the government database about international students — to build a new visuali….

You can now order a nasal spray flu vaccine for self-use. Only costs 9 USD with insurance. Big consumer surplus IMO!

GPT-5 and Claude Opus rejections for bio things are quite over the top right now. I expect there to be more and more complaints from biotech ppl. I appreciate AI/bio risks, but would be good to make rejections more targeted asap.

Great work by Ruxandra. If only 10% of the analytical power levied at FDA reform would be used to improve the European Medicines Agency, I'd be extremely happy.

It'd be great to have something like Launch Sequence for Europe. It ofc would need to be somewhat opinionated; but EU regulation has so many obvious flaws that are worth analyzing and remedying. EU has less progress -> lots of opportunity for progress ppl.

RT @SecureBio: @IFP has released Launch Sequence, a series of essays to advance concrete scientific projects. In it, our very own @Simon__G….

If you want to learn what the piece is about, see the full thread here:

There is much more detailed work to be done to make this type of system a reality. If you want to learn more, feel free to DM. At NAO, we have further write-ups on metagenomic disease surveillance: Full post:

Finally, we need to spur the development of AI-based analytics. The Defense Innovation Unit should thus fund organizations to develop AI-based threat detection methods that enable analysis of large amounts of metagenomic sequencing data.

To make this data usable, HHS should provide guidance on data sharing. This guidance should set clear guidelines on how human material can be removed while guaranteeing rapid data upload and provision of adequate metadata.

Specifically, CDC should allocate $80M to deploy metagenomics across three existing surveillance systems:. - Traveler surveillance ($44 million). - Airplane wastewater surveillance ($10 million). - Clinical Respiratory sample surveillance ($26 million)

How can we deploy metagenomics for pathogen early detection?. CDC might soon receive funding for the launch of a Biothreat Radar Detection System. This new system’s centerpiece should be an ambitious metagenomic sequencing effort.

Unlike targeted approaches, metagenomic sequencing reads out any genetic material, including known and unknown pathogen genomes. Until recently, the routine use of metagenomics was too expensive. However, sequencing has gotten cheaper, and is now ready to be deployed at scale.

To make AI-enabled pathogen detection possible, we need data that companies, researchers, and governments can experiment with. The government can step in here and help, by providing the data most useful for disease surveillance: . Metagenomic sequencing data.

However, these same AI systems and similar systems trained on biological data can also be used for biodefense, by:. - Predicting if DNA or RNA encodes familiar threats. - Detecting anomalies in large amounts of sequencing data. - Manually analyzing suspicious reads.

1918 influenza alone killed 100 million people. Since then, we have invested billions into infectious disease mitigation but little in early detection. Our lack of early detection is untenable: AI systems are becoming more capable in virology, quickly increasing pandemic risk.

RT @richardfuisz: This is the true metascience revolution we were promised.

One of the main contributions in this piece is a graph of sequencing costs that i) isn't a weird 3D shape and ii) contains data beyond 2022.