bioengineer guy @try_litefold
conserved motif
Joined April 2025
- Tweets4.6K
- Following794
- Followers3K
- Likes34.2K
nabajit (bio/acc) retweeted
The cover is out! The lotus (LC3), floating on lily pads (PI lipids) comes together!
@emboreports
link.springer.com/journal/44…
nabajit (bio/acc) retweeted
We (@VanditPatel002) just completed a "Hello-world" task for Weight-Space Programming and open-sourced it.
Can an LLM figure out what a neural network was trained to do, simply by inspecting its weights without running a single forward pass?
We tried this on a small MNIST classifier with random labelling. The results were fascinating.
A thread 🧵
lfg ✨
Mol-1-constraint is now available to every researcher ^^
88.3% success on Foldbench-AB. Code and weights released: github.com/molence/Mol-1/tre…
Clone it. Build on it. Make it part of your research. Or try it directly at lab.molence.com.
Research should benefit everyone. We’re putting that belief into practice.
3
214
nabajit (bio/acc) retweeted
AlphaGenome 2 is helping uncover the cellular drivers of disease, giving scientists a powerful tool to predict how DNA variants impact specific cell types.
Every cell in our body shares the same DNA blueprint, but different cell types (like heart muscle cells vs. immune cells) read that blueprint differently. This new @GoogleDeepMind model predicts how these specialized cells interpret DNA, turning genetic code into biological insights.
Building on AlphaGenome 1 this makes it significantly easier for researchers to link genetic findings to the root causes of disease in the specific cells where they actually unfold.
nabajit (bio/acc) retweeted
Hillclimbing: Make the model better (best intentions).
Benchmarking: Measure how much better (best effort).
Benchmaxxing: Make the numbers look better (best optics).
Google please bring back Pro, Flash, Flash-Lite. The new naming is confusing and would love if Pro users have access to new models from Day 1
4
248
nabajit (bio/acc) retweeted
Given so much talk on PhD stipends, about time we agree that in India, even "good" RA pay (₹55k/mo = ~$7k/yr) is a quarter of a low-end US PhD stipend.
The arbitrage is so extreme that US PhD students can literally "hire" Indian RAs out of pocket or through small grants.
nabajit (bio/acc) retweeted
A lab gets better when the last experiment changes the next design.
Using LiteFold, we ran a peptide campaign against BHRF1. We got 47% in hit rate.
That measurement is the whole point of LiteFold: your data, your target, the next round smarter than the last.
Lots of blogs and technical reports are on the way!
Biology is hard to generalize. But it is ready to specialize on your data and solve your problem.
Your lab already collects plenty of data. What it lacks are the systems that turn it into a learnable signal: state-of-the-art models, environments and evals, built around your problem.
LiteFold converts your own lab, into an AI learning environment, so you can recursively self improve for your the target you are working or the molecule you are designing.
We call this: Recursive Improvement From Experiment Feedback (RIEF)
Try LiteFold today and contact us if you want to integrate LiteFold to your lab.
nabajit (bio/acc) retweeted
Biology is hard to generalize. But it is ready to specialize on your data and solve your problem.
Your lab already collects plenty of data. What it lacks are the systems that turn it into a learnable signal: state-of-the-art models, environments and evals, built around your problem.
LiteFold converts your own lab, into an AI learning environment, so you can recursively self improve for your the target you are working or the molecule you are designing.
We call this: Recursive Improvement From Experiment Feedback (RIEF)
Try LiteFold today and contact us if you want to integrate LiteFold to your lab.
opus 5.5 lsvp lower than sonnet 5.5 without lsvp is crazy lol
A research question can be completely safe, but mention a pathogen or certain lab techniques, and some AI models will refuse to answer.
Our latest benchmark, BenchBench-Refusals, tests how well leading AI models distinguish harmful from benign biology requests.
See the results: benchling.com/benchbench/ref…
ALT Bar chart titled "Can LLMs identify harmful biology requests?" displaying Macro-F1 scores of various language models. GPT-6.1 Sol leads with 88.4%, followed by Sonnet 5.5 at 85.1%, and Opus 5.5 (LSVP) at 84.3%. Fable 5.1 ranks lowest at 37.9%. The chart shows performance variations across models. Benchling logo is at the bottom right.
2
8
704
nabajit (bio/acc) retweeted
My take on the automated labs for biology discussion:
The hard thing about biology is the biology
it's just ragebait at this point
i hate it how good Opus 5.5 is for frontend and code. if only i could use it for some bio research
3
36
2,009
nabajit (bio/acc) retweeted
Dear early-career researchers,
You don't need to publish in Nature to be a good scientist.
You don't need hundreds of citations to make a meaningful contribution.
And you don't need to become a professor to have a successful research career.
Your work matters beyond metrics.
