English
Ruslan Rust
Report

Ruslan Rust✓

@rust_ruslan

Asst Professor @USC with interest in the brain, BBB, stem cells, AI + what can go wrong after stroke & AD. Alum @ETH_en. Fixing science @ResearchHub

2.1K Following3.4K Followers4.5K Posts
  1. Ruslan Rust✓@rust_ruslan
    Lots of debate this week on whether journals should ban AI in peer review. We actually tested this in our new preprint, and the results were pretty clear to me. For context, I've reviewed for most of the big journals and evaluated 1,000 human reviews as an editor. I love to read and evaluate science. TLDR: From now on I won't do any peer review without AI. For clinical trials, I'd even call human-only review dangerous 🧵
    Download
    5922077868.8K
  2. Ruslan Rust✓@rust_ruslan
    Posts @rust_ruslan
    Time is the part I think people underestimate. An AI review took 1.4 to 16 minutes. Human reviewers report 5 to 8.5 hours, and then the paper waits. In our prior study on biomedical preprints, the median time from preprint to journal was about 7 months, and 4% took more than 2 years ( That delay has a cost, especially for clinically relevant work. When RECOVERY found that dexamethasone saves COVID patients, the result went out right away. One analysis estimated that waiting about a month for the journal paper would have cost more than 200 lives in the UK. An AI check could be posted the day a preprint goes up.
    Download
    26242.1K
  3. Ruslan Rust✓@rust_ruslan
    Posts @rust_ruslan
    Some are concerned about confidentiality. My perspective here is that once a manuscript is preprinted, confidentiality is not an issue. For unpublished work, open-weight models you can run on your own computer found up to 89% (Kimi K3). Cost-effective models do well too. GPT-6.1 Sol found 94% of major errors for about 10 cents per review, and GPT-6 Luna 86% for under 3 cents. So my position: for clinical trials, the review we should stop accepting is the fully human one. Journals that ban AI in review are choosing the 27% and are basically not detecting 3 out of 4 major errors. Preprint: (v2 with all 58 trials coming in the next few days! You're getting an exclusive preview here :) w/ @dominikus_brian @p_h_x_i @HaoYin20 @jeffyfish9
    Download
    35352.6K
  4. Ruslan Rust✓@rust_ruslan
    Better cell masks reveal biology missed by zero-shot segmentation. SPInfer raised F1 from 0.37 to 0.84 in human brain images; fine-tuned masks uncovered AD-pathology associations across 69 donors. Observational, not causal.
    Download
    0311618
  5. Ruslan Rust✓@rust_ruslan
    Working on cell therapies for the brain, I always wonder how we should replace lost cells, how many need to be replaced and what happens functionally. Almost impossible to dissect in a mouse. I think this fly model is such a cool way to test it. What if we replace a cell with the wrong cell type? Or replace only 5%? Almost unlimited questions we can ask here
    Download
    Ruslan Rust @rust_ruslan
    Ever wondered if we could simulate a brain injury and predict which cells need to be replaced for recovery? In our new preprint, we tried this in the full fly connectome. Injury site and size mattered a LOT, and so did the identity of the replaced cells. After losing 2,000 neurons, the right 100 (5%) restored 62.5% of flight function 🧠🪰🧵
    117544
  6. Ruslan Rust✓@rust_ruslan
    Ever wondered if we could simulate a brain injury and predict which cells need to be replaced for recovery? In our new preprint, we tried this in the full fly connectome. Injury site and size mattered a LOT, and so did the identity of the replaced cells. After losing 2,000 neurons, the right 100 (5%) restored 62.5% of flight function 🧠🪰🧵
    Download
    23161K
  7. Ruslan Rust✓@rust_ruslan
    Posts @rust_ruslan
    After a 2,000-neuron lesion the fly reached 6.5% of healthy flight distance. We then replaced 100 of the lost motor-control cells with: • the original cells and wiring: 62.5% • the original cells, but random wiring: 8.7% • cells of the right types, wired by type: 60.0% • cells of the wrong types: 6.7% With 2,000 replacement cells, the right types reached 87.6% and random wiring 14.3%.
    Download
    10036
  8. Ruslan Rust✓@rust_ruslan
    Posts @rust_ruslan
    For cell therapy this suggests that the identity of the new cells and how they wire in matter more than the number of cells. These are predictions from a fly model, and next we want to test them in real flies. 📄 Preprint 🎮 Test it out yourself:
    00021
  9. Ruslan Rust✓@rust_ruslan
    Ruslan Rust
    RT @a16z: Formation Bio just announced a $600M fund aimed at drug development - arguably the most meaningful bottleneck in bringing new medicines to patients. Few large pharma companies have been built in the past century. Even fewer have been founder-led. AI is making drug discovery dramatically more productive, but promising drugs still compete for limited capital, experienced teams, and clinical development resources. The Drug Development Accelerator Fund pairs dedicated capital with Formation Bio’s drug developers and AI platform to advance a larger portfolio of drugs toward patients. Congrats @BenjamineYLiu, @lintaho, and the @formationbio team!
    Ben Liu @BenjamineYLiu
    Today we’re announcing the $600M Formation Bio Drug Development Accelerator Fund. @lintaho and I started @FormationBio because we believed the biggest bottleneck in bringing new medicines to patients wasn’t drug discovery, but drug development. As AI dramatically accelerates discovery, that bottleneck is only becoming more acute. Without the capital and capacity to advance promising drugs through clinical trials, much of that progress won’t translate into medicines for patients. Our new Fund brings together dedicated capital, drug development expertise, and AI-powered development capabilities to acquire promising drugs and advance them through the clinic.
    018012
  10. Ben Liu✓@BenjamineYLiu
    Today we’re announcing the $600M Formation Bio Drug Development Accelerator Fund. @lintaho and I started @FormationBio because we believed the biggest bottleneck in bringing new medicines to patients wasn’t drug discovery, but drug development. As AI dramatically accelerates discovery, that bottleneck is only becoming more acute. Without the capital and capacity to advance promising drugs through clinical trials, much of that progress won’t translate into medicines for patients. Our new Fund brings together dedicated capital, drug development expertise, and AI-powered development capabilities to acquire promising drugs and advance them through the clinic.
    Download
    6972763149K
  11. Ruslan Rust✓@rust_ruslan
    Ruslan Rust
    RT @jrkelly: Awesome to see this. Really excited about the 1-person biotech in particular. CROs with APIs available for agents like @ginkgo’s Datapoints service plus something like a “cursor for biotech” could really change the cost of launching biotech startups.
    Anthropic @AnthropicAI
    Over the past few months, Anthropic scientists have engaged in deep research partnerships with six organizations developing new medicines. We've also begun drug discovery efforts of our own. In doing so, we’ve learned much more about how Claude can support the many stages of drug development. In this blog post, we share some early results from these efforts, and describe how this work fits in with our wider ambitions in the life sciences.
    0409
  12. Anthropic✓@AnthropicAI
    Over the past few months, Anthropic scientists have engaged in deep research partnerships with six organizations developing new medicines. We've also begun drug discovery efforts of our own. In doing so, we’ve learned much more about how Claude can support the many stages of drug development. In this blog post, we share some early results from these efforts, and describe how this work fits in with our wider ambitions in the life sciences.
    6881.1K10.5K3.1M
  13. Ruslan Rust✓@rust_ruslan
    Ruslan Rust
    RT @ResearchHub: 12 people put up $1,400 for four synthetic genes. Two of them encode DNA polymerases reconstructed from bacterial ancestors ~133 million years back from anything alive today. That proposal is now a published Stage 1 Registered Report. No editor approved it. Funding did.
    05010
  14. Ruslan Rust✓@rust_ruslan
    Plasma p-tau217 was most strongly linked to clinical progression in a 3,645-person memory-clinic cohort. Longitudinal analyses also supported GFAP and NfL. Population-specific thresholds still need validation. Preprint:
    Download
    107228
  15. Ruslan Rust✓@rust_ruslan
    Posts @rust_ruslan
    The identity of the new cells also mattered. When they wired in by cell type, 100 cells of the right types reached 60% of healthy distance, the wrong types 7%. The right cells with random wiring: 9%. For stem cell therapy after stroke this is probably the most relevant part.
    Download
    10069
  17. Show more posts