Tom Ellis

Biolab robots have been a big topic on here this week. Not everyone $2m+ to spend, but plenty of labs have an OpenTrons and we're happy to now share a standard Golden Gate Cloning Workflow for their latest models, that we call "SLOWPOKE" pubs.acs.org/doi/10.1021/ac…

@SynBio1 If my attempts are anything to go by they probably packed 30 jokes in but the audience only spotted 19 of them. I think I attempted 15 jokes in my SEED 2024 keynote, and only remember laughs and groans about half the time. Tough crowd

Shout out to the absolute legends who somehow packed 19 jokes into their conference talks as documented in this study about the use of humor on scientific talks royalsocietypublishing.org/rspb/article/2…

@ProfTomEllis Gemini Claude Doubao( from ByteDance, owns TikTok), Yuanbao(basically DeepSeek R1)

@FrostJiangShan I think any list that puts affinity above stability is wrong so I'll give this as a win to Gemini. But this is just my personal take

Okay here's a game - you're making nanobody therapeutics. AI can help you optimise them, but you need to prioritise order of feature importance. What is your order? Affinity Specificity Expressibility Clinical toxicity In vivo stability Purified stability Patentability Other?

@SynBio1 @JoeMcAuliffe17 I should've been clearer in the original tweet but in this thought game of mine, the available AI models are all equally good at doing their tasks, but you need to provide the priority order. The model's ability gets worse if they are low priority as design space gets less.

@JoeMcAuliffe17 @ProfTomEllis This is the most AI-ready problem on the list though

@DrTregoning @ATinyGreenCell It's editor-enforced based on Cell's policy of shrinking supplementary materials into a low number of figures after the paper is accepted.

@ATinyGreenCell editor/ reviewer fail for letting this through

Bruh. This is why supplemental materials section is a thing. Don't do this to your readers, please.

@ATinyGreenCell @Nicole_Paulk because after acceptance of the paper, Cell suddenly asks you to limit the supplementary materials to being no longer than 8 supplementary figures in total. So authors are at the last minute asked to combine all their Sfigs into multipanel figures. It's a totally stupid rule.

@Nicole_Paulk wild work to not split this up lol whygodwhy

@Banana_Oncology As the only people who care about this are the copy editing admins, it’s a totally natural response for the authors to go ‘whatever let’s just stick all of figures into the biggest multi panel figure of all time and get this over the line’

@Banana_Oncology They only ask authors to do this after peer review is done so you end up in a situation where you can’t NOT show all the supplementary figures (as they were part of the accepted paper) and somehow you have to go from 30+ SFigs to only 8 SFigs without changing the paper.

Ok this figure is pretty intimidating...

@merichyoztyurk Definitely last

@ProfTomEllis doesn't patentability depend on all? 😂

@ProfTomEllis Specificity and toxicity first, always. No therapeutic survives without safety. Affinity and in vivo stability next for efficacy. Expressibility matters for scale. Patentability comes last.

@ProfTomEllis I'd prioritize ADMET (Clinical toxicity) because it is the hardest to get with AI alone. If I had some kind of tool that could predict ADMET reliably, I'm confident you could get the other properties pretty good with existing tooling.

@SynBio1 For me expressability is top because if it's not quick and easy to manufacture then it bottlenecks everything that requires a real-world test or trial and hits speed and profit badly.

Awesome new synthetic biology paper in Nature today led by Bolin An (SIAT) and Zijay Tang (Harvard/MIT) on cell fate gene circuits that specialise cell progeny to different tasks and control their population ratios. This is the most #synbio paper ever! nature.com/articles/s4158…

Engineering E. coli for controlled growth: Instead of exponential, scientists achieved linear proliferation, improving safety in therapeutic applications. Thoughts? PMID:41723156, Nat Commun 2026, @NatureComms doi.org/10.1038/s41467… #AI #Pharma #BioMed #RNA #ASHG #ESHG

Have you ever wondered why children starting nursery pick up so many germs? I certainly did, and with a crack team of parent-scientists/clinicians, we set out to answer that question. Is childcare a germ factory or an immune bootcamp? journals.asm.org/doi/10.1128/cm…

Earth's toughest microbes could help save the planet and find life on other worlds share.google/3vpVrsv4r3Zmey…

Whoa. Never had a tie before. I guess I’m going to have to buy another prize. Congrats to Tinkerbell and the Fox from Zootopia. Both amazing bioart submissions!