🔥The 9th Round of Easy Loan, Earn $40 Reward is in progress❗️ ⏰ Promotion Period: January 15th - Feburary 15th, 2025 👉 Register now and check more details at gate.io/campaigns/358

New review out today in Nature Biotech on nanopore-based protein sequencing. Feels... like... something... coming... soon... 👀

Of >105,000 participants with 30-year follow-up, only 9.3% achieved healthy aging (age 70, w/o any chronic diseases). Their diet was significantly associated with this outcome🧵 Nature Medicine

Small proteins can be more complex than they look! We know proteins fluctuate between different conformations- but how much? How does it vary protein to protein? Can highly stable domains have low stability segments? Állan Ferrari experimentally tested >5,000 domains to find out!

Our pLM-designed peptides, when attached to E3 ligases (aka uAbs), enable us to programmably degrade disease targets (think CRISPRi, but for proteins). But what if we want to STABILIZE useful proteins? In Nature Communications, we introduce duAbs! 🌟 📜: nature.com/articles/s4146… 🧬:

I really like this ProGen3 paper because, contrary to the title, I think it actually shows there is relatively little to be gained from massively scaling protein language models. 1/n

Has anyone else noticed that the quality of ChatGPT has gone down significantly recently? Spend more time correcting it than using it. OpenAI I need my money back.

Read more: Predicting pose distribution of protein domains connected by flexible linkers is an unsolved problem biorxiv.org/content/10.110…

Flow matching produces smooth, deterministic trajectories. In contrast, the sampling process of a diffusion model is chaotic, resembling the random motion of gas particles.

#iampluto

new paper from our work at Meta! **GPT-style language models memorize 3.6 bits per param** we compute capacity by measuring total bits memorized, using some theory from Shannon (1953) shockingly, the memorization-datasize curves look like this: ___________ / / (🧵)

A really great case study on how everyone wants to train flashy deep learning models but nobody wants to spend the time cleaning/curating the datasets and inspecting the quality of model predictions. This is very prevalent in protein AI research. rachel.fast.ai/posts/2025-06-…

Gaia: An AI-enabled genomic context–aware platform for protein sequence annotation Science Advances １．Gaia introduces a new paradigm in protein function annotation by combining sequence, structural, and crucially, genomic context information into a unified, scalable search

It is imperative that we continue to fund CASP and other similar competitions that fuel innovation in computational protein science and underpin future biotechnology innovations. Please share and reach out to your congress representatives! #FundCASP science.org/content/articl…

Ambient Proteins: Training Diffusion Models on Low Quality Structures 1. A new framework, Ambient Protein Diffusion, revolutionizes protein structure generation by leveraging low-confidence AlphaFold structures as valuable, corrupted training data instead of discarding them.

AI just shrank evolutionary analysis from hours → heartbeats🧬 SelRegAA transformer classifies evo selection for amino‑acid sites ≈1,000× faster than codeml ⚡️ DL models are replacing classical likelihood methods in molecular evolution 😅

Had a lot of fun learning diffusion and addressing key issues in protein diffusion with Giannis Daras Jeffrey Ouyang-Zhang TLDR: a few protein structure insights inspired us to design a new diffusion loss, training regime and dataset, resulting in significant performance improvements

That prompt is so accurate 🤣🤣 Got so many try except hacks to make sure the LLM returns a valid JSON string.

Toxic heavy metals contaminate ecosystems and harm vulnerable communities. 🌎⚠️ Here, we present Metalorian, a conditional diffusion model that generates heavy metal-binding peptides, validated by MD and experiments! 💻🧪 📜: biorxiv.org/content/10.110… 🤗: huggingface.co/ChatterjeeLab/…

Turns out using co-pilot/AI during software development actually slows down productivity in a randomized clinical trial sitting. 🤔 metr.org/blog/2025-07-1… Personally, it saves me a lot of time when writing data analysis code, or refactoring bad code. What do y’all think?

AlphaFlex: Accuracy Modeling Of Protein Multiple Conformations Via Predicted Flexible Residues A new computational framework, AlphaFlex, has been introduced to address the long-standing challenge of accurately and efficiently modeling multiple protein conformational states, a