"A single binary operator, eml(x,y) = exp(x) − ln(y), together with the constant 1, generates the standard repertoire of a scientific calculator"
https://t.co/UJtSFjvEhD
💡 Did you know that #LightDock was developed by Zymvol’s CTO @br_jimenez , together with @dxbarradas and @jroeltou ?
If Structural Biology is your thing, try out this tool for modelling interactions between macromolecules.
✅ Free
✅ Open server
✅ AI-powered
Excited to share that our recent research using #Martini3#CG simulations on the role of Spc2 in the yeast signal peptidase complex (SPC) has been published in the @JCellBiol
Read the full article here: https://t.co/fsSgaTzPmI
Hace unos días hablaba con un grupo de personas allegadas sobre si dejar Twitter o no. He estado a punto. Pero al final decidí quedarme precisamente por lo que dice Alfonso. Me niego a ceder… Por mucho que sepa que probablemente no sirva para nada.
Could the “guy who worked on Alphafold” saying “This paper sure saved our ass on that Nobel prize” come forward please? Maybe a citation would have been nice 😪.
The failure of authorities to issue public warnings in time before the devastating floods in Valencia offer an example of what it means when far-right policies meet environmental disasters.
One of Vox's conditions for backing the PP's Carlos Mazón as regional premier was the elimination of the Valencian Emergency Response Unit.
I can't stop thinking about this blog post where they replaced Redis with SQLite—and surprisingly, SQLite was faster!
What's interesting is that Redis was running locally, SQLite was storing the data on disk. So it was memory (Redis) vs disk (SQLite), but Redis needed to communicate through IPC.
If you're looking for failure in AI-driven (or any) drug discovery, you'll easily find it.
The hard work is properly benchmarking various approaches for relevant tasks and publicly reporting your findings fairly.
Where is AI winning?
https://t.co/Cp84xdoxbx
WOW. @Microsoft just open-sourced the code for one of "THE MOST" influential Paper of 2024 🔥
1-bit LLMs (e.g., BitNet b1.58).
Now you can run a 100B param models on local devices quantized with BitNet b1.58 on single CPU at 5-7 tokens/sec 🤯
The dream we have all been waiting for.
📊 Performance Improvements:
- Achieves speedups of 1.37x to 5.07x on ARM CPUs
- Larger models see greater performance gains
- Reduces energy consumption by 55.4% to 70.0% on ARM
- On x86 CPUs, speedups range from 2.37x to 6.17x
#HIRING We're looking for a #MolecularModeler to help us advance our transformative technology to multiple on-going projects.
Check out all our job openings at: https://t.co/fdoBktXLe6
Llevo casi 50 plazas universitarias a las que me he presentado en los últimos años y, básicamente, 9 de cada 10 son así. Es de vergüenza como se gestionan estas plazas públicas y se nos expulsa del sector universitario a quienes no tenemos un padrino o departamento detrás.
@hokru_science Not trying to ashame anyone here. One thing is sharing code and polishing/improving it and then publishing. This approach is nice and you can benefit from early sharing from the community. But publishing a manuscript with such low quality of the code makes me unsure of results.
Can we stop publishing software in science without testing it properly? Having a couple of examples and 2 notebooks in the repository is not testing. FGS, it hurts even more when the group is from a technical university.
@_inc0_ @NGSorcerer As there is no novelty in running triplets of your experiments at the wet lab, same applies for code. This is about reproducibility and repeatability. If you don't store your random seeds, how am I going to reproduce your results and trust your science?
@hokru_science I'm sorry to disagree: I don't share the illusion of "at least it's published". The same way you run triplets in the wet lab, you test your code. Reproducibility is a must and a basic pillar in science. You teach yourself diffusion for 6 months, but not CS...
Set and store your random seeds, use proper logging and not prints, don't write 400 LOCs nested on a while True. Don't write redundant functions for wrapping a single line call with the same function parameters. I stop here, I need a break. Follow me for more code review tips 😂
Make it be a Python package. Write docstrings, lint the code. Teach yourself some OOP, don't add binaries to the repos (WTF), don't "import *". Don't use all available resources (GPUs detected, all for me now)...
There is a new tool on BioIcons to turn any protein (in PDB format) into a 2D vector illustration in SVG format. You can upload any PDB file or import from PDB, AlphaFoldDB or ESMFoldDB. Feedback welcome :)
https://t.co/grMysO8JV5
🎯 Job Alert: #AI Scientist!
💡Are you a talented AI Scientist? Come work with us in #Barcelona to build the future models at the interface of AI and #enzyme discovery & engineering!
We offer a salary of €45k-60k, but we are open to discuss with more experienced candidates! 👇🏽