Oscar Palomino

📣 Hiring a postdoc! 📣 We are looking for a postdoctoral researcher in molecular simulations in Johannes Gutenberg Universität, Mainz, Germany. Please share! For details, visit the link below. euraxess.ec.europa.eu/jobs/312665

@PetarV_93 Dear Petar, this sounds very nice, thanks for sharing! One question: is there info about the lectures that will be covered?

~3 days left to apply to attend EEML'26 in Cetinje 🇲🇪 (link below). We've faced unprecedented levels of interest and therefore will **not** be extending the deadline after the 31st. The speaker lineup is stellar, atmosphere will be amazing; you don't want to miss this one :)

Hot take: it's 2026 we don't need to be manually doing citations or asking llms to do them overleaf should just autogenerate the citations from dois

@biocheMichael Can I ask RMSD of what?

Here is a view of some of the minibinders and scFvs I have produced using Boltzgen. I have compared the residue count to the iPTM and the color scale represents the pae. You can see that Boltzgen generates a lot of different types of structures. There does not seem to be a bias towards 'high quality' and it is purely a numbers game. This is why I am working on a set of experiments which I hope to use the information learned to push these results directionally in my favor.

@AriWagen Dear Ari, thanks for the post! 1) will there be another one for TS opt benchmarking? And 2) is there any code associated to this benchmarks? Thanks!

I ran the same 25 optimizations with every combo. of four optimizers (Sella, geomeTRIC [both "cart" and "tric"], ASE/FIRE, and ASE/L-BFGS) and four NNPs (OrbMol, OMol25's eSEN sm. cons., AIMNet2, and Egret-1) & GFN2-xTB. rowansci.com/blog/which-opt…

@eigensteve I would be happy to: I am teaching also a course on similar lines! Would be nice to see other fresh perspectives!

New Book & Video Series!!! (late 2025) Optimization Bootcamp: Applications in Machine Learning, Control, and Inverse Problems Comment for a sneak peak to help proofread and I'll DM (proof reading, typos, HW problems, all get acknowledgment in book!)

@gregoria_i_ Procesos de que? :v

No supe decir que no y terminé en estudiando procesos en el Instituto de Física 🙃

@jonkhler Do you think that the protests help shift the general view? Thinking on things like Fridays for future which I am also not sure of the impact on that. It is unfortunate how the general view is shifted now more to the right, but I see no way out now, sadly.

Germans are simply embarrassing. 🤦‍♂️ After the conservatives nearly passed unconstitutional nonsense law using right wing talk and the votes of the nazis resulting in mass protests all over the country: the voter just doesn't care... You vote what you deserve. 😓

Remember when LLMs used to make mistakes?

@camhowe1729 @francoisfleuret Assume that you are sampling from an area: the area between the x axis and the function. Ofc, if the distance between them x axis and the graph is higher might mean you get more chances of sampling from that value. Does this make sense?

@francoisfleuret yeah but I'm struggling to get how uniform randomness changes into the distribution's randomness.

@alfcnz I mean, I remember that there was a time that I explained the results of concatenating neurons as when a cow has to digest the food multiple times. Kinda works, but I also don’t want to say that in class xD

@opehache Yeah, it allows for an intellectual playground to try out explanations before delivering them in class. 😀😀😀

Fuck it, I’m going back to mentoring students again. Three years ago, when I started writing my book, I decided it would be my only disciple, so that many would benefit from reading it. Now that I need to create more edu content, I’m going to seek help from my new mentees.

“maybe I’ll do some writing between christmas and new year” Between christmas and new year:

Spending an hour reading papers vs spending 10 hours perfecting my figures

Google has made us incapable of memorizing complex things by reducing memorization to search queries. LLMs will make us incapable of solving complex problems by reducing problem-solving skills to writing prompts.

Here's a first: I'm hiring a PhD student at @WUR. If you have (or know someone who has) a strong background and interest in computational (bio)chemistry or computer science (especially 3D computer vision), let's work together 😊 wur.nl/nl/vacature/ph…

@DdelAlamo MD is molecular docking, right? :p

Papers combining MD & ML (MD is the gun)

PAPER DEADLINE IS NOVEMBER 1st!!! ELLIS Machine Learning for Molecules Workshop 2024! Submit your papers here: openreview.net/group?id=ELLIS… Workshop website: moleculediscovery.github.io/workshop2024/

@DdelAlamo @inve_michele @PaulRobustelli In my experience these type of collapses does not go easy, even with IDP specific force fields. I would suggest either OPES with energy biasing or HREX

@inve_michele @PaulRobustelli Nope, there's a folded domain involved. Obv can't get into specific protein but I can't chop it off

Question for folks running MD on IDPs. Starting from spaghetti AF models len 50-150 residues, they nucleate/scrunch up into compact structures quickly, in 10s of ns. Is this normal/expected and/or protein-specific?? Are they ever expected to un-scrunch w/o enhanced sampling?

@skepteis Was this prophetic?

One of the absolutely last things that science needs is to boost the egos of (already successful) scientists by giving them prizes.