The Round-Up #4
AI-designed proteins, embryo editing, Coruscant IRL, and more!
Welcome, or welcome back, to The Science Fictional Now! Today’s post is the latest in our Round-Up series, featuring a selection of the most interesting and informative things I’ve read (or listened to/watched) over the past few months. This time around, we’ve got science news ranging from new embryo editing companies to advances in AI-powered protein design and projects to detect the spread of pandemic viruses. If you’re in a more interstellar mood, we’ve got a great video on the feasibility of making a planet-cities like Coruscant in Star Wars.
As always, let me know which stories you found the most interesting in the comments below. Enjoy!
Contents
Advances in protein design
Per usual, we’ve got some interesting new scientific papers to cover…but first, a note on terminology! “Protein engineering” refers to the practice of changing proteins’ amino acid sequences to make them better at their jobs in some way: more efficient at catalyzing a chemical reactions, able to do so at higher temperatures, etc. “Protein design” essentially refers to the same process when done using computers, especially with AI/machine learning-based programs. As its name implies, protein design promises more drastic changes to protein’s sequence, structure, and even function than traditional engineering efforts. Both of the papers discussed below are in this area of study.
Superstable proteins: Peng Zheng’s lab at Nanjing University recently published a paper in Nature Chemistry redesigned proteins found in muscle fibers using AI, making them several hundred times more resistant to force and able to keep their shape up to 150 ºC. This type of heat resistance is reminiscent of infectious proteins called prions (which most notably case mad cow disease), and the Zheng lab’s work sparked biosecurity discussion online. I’m personally more excited than worried about this work since (1) it doesn’t relate to prions’ actual mode of infection (2) natural prions can already survive temperatures way above 150 ºC. Check out David Manheim’s X/Twitter thread on the paper for another more fleshed out perspective.
AI-designed genome editors: Our second protein design story involves the design of small genome editing proteins…and also happens to be gratuitous self-promotion. This project occupied most of my time in Jennifer Doudna’s lab at UC Berkeley and I’m glad to see it out in preprint form. Here’s the summary I wrote for LinkedIn:
In the paper, we used machine learning to design a set of small RNA-guided nucleases with diverse sequences, all based on a CRISPR-Cas12 ancestor called TnpB. The minimal size of these complexes makes them ideal for delivery into human and plant cells, and after screening activity in bacteria, we demonstrated efficient genome editing in both of those systems. We also solved the first experimental structure of an AI-designed genome editor, which revealed that our protein contacted its target DNA and guide RNA (which specifies the sequence the editor cuts at) in different spots from our base enzyme.
For a bit more scientific detail, check out these X/Twitter and Bluesky threads describing some of the key experiments we did!
Embryo editing companies
We live in a time where it’s not only possible to literally edit human DNA using technologies like CRISPR, but medicines exist to do this very thing. Crucially though, these medicines only change cells in patients’ bodies - not the germ cells that will go on to be their children. In 2018, when biophysicist He Jiankui announced the birth of twin girls from embryos he had modified, he was condemned by the scientific establishment and spent three years in prison.
Embryo editing has been a taboo topic among scientists and clinicians since. Legally however, it exists in a grey area. The act of editing an embryo is not itself illegal, though such research cannot be done with federal funds. Privately-developed embryo editing therapies would need to be approved by the FDA. Recently, two companies have emerged hoping to develop this type of technology.
Manhattan Genomics: Back in August, biotech entrepreneurs Cathy Tie and Eriona Hysolli announced the formation of Manhattan Genomics,1 with the goal of editing embryos to prevent genetic disease. Tie went on Ashlee Vance’s Core Memory podcast in September, WIRED ran a story on the company in October, and in November, ethics advisor John Quain led publication of a preprint suggesting situations where germline editing should be allowed.
Preventive: At the end of October, genome editing entrepreneur Lucas Harrington (a fellow Doudna lab alum) announced a new embryo editing company called Preventive, which a Wall Street Journal article later revealed was backed by tech billionaires Sam Altman and Brian Armstrong. Katherine Long, who co-wrote the piece, later appeared on a KQED panel with Berkeley’s Fyodor Urnov and the Center for Genetics and Society’s Katie Hasson to discuss embryo editing.
Tie has talked about wanting to keep everything at her company open and above board, while Harrington has stated that Preventive’s goal is to answer the open-ended question of whether embryo editing can be done safely. Data scientist Stephen Turner explicitly stated that he got involved with Manhattan Genomics to ensure research was done transparently and with a medical focus. Harrington’s previous comments suggest he’s more or less in the same boat. I mention all of this to give readers confidence that we at least won’t live in Brave New World/Gattaca/God Emperor of Dune starting tomorrow. There are obviously lots of issues to discuss here though, and I’ll likely write a dedicated post on embryo editing before too long.
Science and government
Experiments in science funding: Last week, the National Science Foundation announced a new program called Tech Labs, with will issue large long-term grants to researchers at non-university labs. Caleb Watney of the Institute for Progress (nonpartisan DC think tank) laid out the program’s vision in a Wall Street Journal op-ed at X/Twitter thread. See also these two threads from Anshul Kundaje for additional info and food for thought.
Monarez’s vision for public health: Back in August, Susan Monarez was ousted from the CDC after less than a month as its director, having clashed with RFK, Jr. over vaccine policy. Her first interview since the incident came in November, when she sat down with Eric Topol; you can watch the video on YouTube or Substack. I especially recommend listening to the section from ~33 minutes in, where Monarez reflected on her time at the CDC and discussed her technologically savvy vision for public health.
I’m definitely behind giving bigger, longer grants and funding new types of organizations, but I do have some lingering questions about the initiative. Why focus exclusively on seeding entirely new organizations? And why can’t academic researchers also apply for these grants, especially since many of the model organizations mentioned are entirely made up of university labs? Is tech transfer readiness really the only metric grant progress will be assessed on?
One answer to all of these questions is that such program specifications fit with the Trump Administration’s priorities: woke universities don’t deserve federal dollars and science is only valuable insofar as it bolsters the economy. I would be open to hearing more policy-based justifications nonetheless.
Finally, I should note that this isn’t the most opportune time for funding experimentation as the NSF’s 2026 budget is projected at just $3.9 billion, down from $10.2 billion this year. Watney’s op-ed states that Tech Labs will dispense up to $1 billion over the next five years; $200 million would be just over 5% of the NSF’s 2026 budget. That doesn’t seem totally unreasonable (and I’m not really knowledgeable enough to judge) but I’d be curious to know how this figure was reached.
ACIP nixes hepatitis B birth requirement: Beginning its review of the US child vaccination schedule, the CDC’s vaccination panel voted to no longer require compulsory hepatitis B vaccination at birth. Professional societies, local health officials, and individual hospitals decried the move, highlighting just how inconsistent health recommendations have become within the US.
Genome editing clinical trial models: In June’s Round-Up, I mentioned the first personalized genome editing therapy, created for “Baby KJ” Muldoon, who was born with a severe metabolic disorder called CPS1 deficiency but recently took his first steps. As exciting as it was, the episode begged the question of how subsequent genome editing treatments might scale. UC Berkeley’s Fyodor Urnov, whose lab helped develop Baby KJ’s treatment, recently wrote about new clinical trial models that could help this happen.
Biosecurity
Low-tech biosecurity: On a recent episode of the 80,000 Hours Podcast, Andrew Snyder-Beattie of Coefficient Giving (formerly Open Philanthropy) described an interesting new biosecurity initiative called the “Four Pillars”. It’s billed as “low-tech” and is focused on non-pharmaceutical interventions like personal protective equipment, building controls, and metagenomic sequencing.2 I especially like that the plan is set up to be implementable by a philanthropic organization like Coefficient, given the lack of biosecurity voices in the federal government. The first part of the interview also features nice explanations of key biosecurity concepts, including mirror bacteria, the USSR’s bioweapons program, and certain AI bioweaponeering scenarios.
Outbreak early warnings: Last month, the MacArthur Foundation awarded $100 million to the Sentinel project - not the “let’s modernize the US nuclear arsenal” Sentinel project - the “let’s use large-scale pathogen detection to head off future pandemics” Sentinel project (phew!). The project is run by Harvard’s Pardis Sabeti and Christian Happi of Redeemer’s University in Nigeria, both of whom were featured in this 2023 NPR article on the project. The grant marks a crucial and well-timed investment in global health and pandemic prevention.
Bird flu in Antartica: Rene Ebersole had a nice article in this month’s National Geographic, detailing molecular ecologist Jane Younger’s quest to track H5N1 and other pathogens to remote areas of the Earth. This is another very interesting type of disease surveillance that has the potential both to aid animal biosecurity efforts and project global warming’s influence on the spread of disease.
Other fun stuff
Coruscant IRL?: Dami Lee recently put out a great YouTube video on the feasibility of making a planet-wide city. Think Coruscant from Star Wars or Trantor from Asimov’s Foundation. Spoiler alert: this would be extremely hard, but people have thought about it and generated interesting urban planning ideas in the process. Dami has lots of great videos looking at SFF architecture and urbanism and I highly recommend her channel!
Charts for the masses: Saloni Dattani (formerly of the ineffable Our World in Data) wrote a great guide to data visualization over on her Substack. It’s super helpful. Go give it a read!
At the Mountains of…Germany: Xander Balwit wrote an article for Asimov Press on Arabidopsis thaliana, tracing its journey from the Harz mountains of Germany to plant biology labs all over the world. I’m of the opinion that model organisms aren’t discussed nearly enough in popular science given how much research they’ve shaped, so I’d encourage everyone to read the piece. Scientists have developed all kinds of new tools for use in specific model organisms and cell lines, making them increasingly advantageous to work with. Our paper on AI-designed TnpBs above used three such model systems: E. coli bacteria, Arabidopsis, and HEK293T human cells.
Arabidopsis up close. Source.
There’s some stuff going on here that weirds me out. Tie had this weird public romance with He, the company was originally called the Manhattan Project (like, the one that developed nukes), etc.
It also advocates broad-spectrum antivirals, which I’d argue are quite high-tech, but that’s ok.
Great post! Coefficient Giving is the new name of Open Philanthropy ;)