Hiren Joshi

I could be mistaken, but they seem to be stabilising the structure (disulfide bonds?). Folding catalysts, if you will.

This is great, can’t wait to try it out on some stuff I tried using FoldSeek ages ago!

I learnt today that a hallmark of the LLMs is correct usage of the m-dash, a form of punctuation that I regularly and stubbornly use — now turned against me to forever tarnish my reviews as AI generated.

Kim Jong-un hails new North Korean beach resort as one of country’s ‘greatest feats’ this year North Korean leader was accompanied by his daughter Kim Ju-ae, widely presumed to be his heir, at opening of Wonsan Kalma tourist zone

The link fell off the post www.theguardian.com/world/2025/j...

Kim Jong-un inspecting the latest in shark tank technology, that looks suspiciously like a water slide.

Kim Jong-un inspects the latest in North Korean shark tank technology.

If anyone is keen on testing the reglycosylation plugin for PyMOL, please get in touch! There’s a few rough edges I’d like to iron out, and would appreciate feedback.

This Lorenz Oken guy might just be on to something. The article mixes slime and mucus interchangeably, but it is a good read about the relative rarity of complex life. www.asimov.press/p/slime

CCG (incl @manno.se and @xylobecs.bsky.social) representing at the half marathon across the Øresund bridge today. Lovely weather for it!

I already love using it. You select a bit of the disordered protein, run the plugin, add the glyco, and then repeat - so you can build up the full structure piece by piece. Totally worth the effort!

It was pretty easy to create a version that spoke the IUPAC dialect being used in the API. Ojas has been mega helpful in getting this all working. Now it needs a bit of testing and polishing before we can send it out into the world!

You'll notice the SugarBuilder widget in there - it's been extended to work essentially as a <select> element for a set of sugars. I've re-tasked the biosynthetic pathway awareness, to make it so you can only draw the sugars that ReGlyco supports (kinda), so it works as a selector.

Well, it turns out that you can actually make PyMOL do quite a lot of stuff. Here's the full workflow, integrating the amazing ReGlyco into PyMOL to re-glycosylate the LDLA linker. There's some serious magic needed to get it running as fast as it is within PyMOL. @elisafadda.bsky.social

Glycosylation of Structured Protein Domains in Cell-Free Reaction Environments The production of N-linked glycoproteins in genetically tractable bacterial hosts and their cell-free extracts holds great promise for low-cost, customizable, and distributed biomanufacturing of glyco...

Glycosylation of Structured Protein Domains in Cell-Free Reaction Environments | ACS Synthetic Biology pubs.acs.org/doi/10.1021/...

Interesting thread below! One thing I have been trying to understand is how NNs can store so much info in such a relatively “small” parameter space, such that the concepts are stored in the “direction” in a latent space. This idea would fit nicely with this model.

Slime An excerpt from Slime by Susanne Wedlich, translated from the German by Ayça Türkoğlu.

Looks intriguing! granta.com/slime/

On the Biology of a Large Language Model We investigate the internal mechanisms used by Claude 3.5 Haiku — Anthropic's lightweight production model — in a variety of contexts, using our circuit tracing methodology.

This looks like a great resource to gain a conceptual understanding of deep learning models. It would be amazing to have a “Design Patterns” style book a la Gang of Four, just with patterns to detect, and the method to train it.

Congrats!

📌

Systematic and comprehensive analysis of major localizations of alpha-dystroglycan-specific modifying enzymes Abstract. DG binds to extracellular matrix via its O-glycans, which are sequentially modified in a specific order by DG-specific enzymes: POMGNT2, B3GalNT2

Systematic and comprehensive analysis of major localizations of alpha-dystroglycan-specific modifying enzymes
#glycotime #Glycobiology

That, and the Windows API dependencies. Would love to know what the reasoning is there. It isn’t like MS users are not tolerant to expert level GUIs, so I don’t buy the interface argument.

I would like to understand the C# usage in mass spec land. Is this because vendors (Thermo?) have developed their stuff in C#? Some specific libraries only available in C#?

Who knew that AUKUS cares about more than just marine craft.

Human coronavirus HKU1 spike structures reveal the basis for sialoglycan specificity and carbohydrate-promoted conformational changes pubmed.ncbi.nlm.nih.gov/40324974/ #cryoEM

Yeah, would be ideal if they used the colours, but at least the shapes are right!

It’s desaturated as a conscious strategy to make the figure less engaging, otherwise people wouldn’t move past the figure?

A little bit of glycan binding structure for your Tuesday.

Bad year to be a right wing opposition leader in the commonwealth.

Obviously, bad. But doubly so because I would actually be really interested in seeing a galectin / Tim-3 interaction structure!

This issue has looked a lot like people arguing about LLMs and the pros/cons on society.

I wonder if anyone has worked out the logistics/efficiency of running a classifier over posts to search for on topic posts. I guess it can be batched and doesn’t need to be real-time.

With all the standard reservations about what the models can do, or whether evolution has had enough time/challenges to instantiate a version of a protein for all functions.