Kat

131 posts

Kat banner
Kat

Kat

@katyenko

building @muni_bio AI ∩ bio | in vitro to in silico, always an experimenter

SF Katılım Ağustos 2024
488 Takip Edilen544 Takipçiler
Sabitlenmiş Tweet
Kat
Kat@katyenko·
The future we bet on at @muni_bio is one where agentic workflows become the default in early drug discovery. In our latest post, we walk through a pipeline designed and executed autonomously using @RowanSci tools for our TBXT hackathon. These candidates were submitted to @onepot_ai alongside compounds designed by other human participants. As agentic loops continue to reshape every part of the drug discovery pipeline, scientists and engineers need to think about grounding their in-silico loops in the wet lab to avoid lead-slop. muni.bio/research/muni-…
English
3
10
62
6.3K
emma
emma@emmamariecryer·
after months of plotting, i have finally convinced our team to start an @OpenRouter Chess Club. i am considering this the highlight of my career. first move in SF - see you August 12th!
emma tweet media
English
7
1
17
2K
Kat retweetledi
Corin Wagen
Corin Wagen@CorinWagen·
I'm a big fan of how Kat and Derek are working to make small-molecule design and simulation, a field with a lot of unindexed tacit knowledge, into legible workflows that can be studied and optimized. In this post, they lay out their full pipeline for TBXT binder design:
Kat@katyenko

The future we bet on at @muni_bio is one where agentic workflows become the default in early drug discovery. In our latest post, we walk through a pipeline designed and executed autonomously using @RowanSci tools for our TBXT hackathon. These candidates were submitted to @onepot_ai alongside compounds designed by other human participants. As agentic loops continue to reshape every part of the drug discovery pipeline, scientists and engineers need to think about grounding their in-silico loops in the wet lab to avoid lead-slop. muni.bio/research/muni-…

English
0
3
15
2.6K
Kat
Kat@katyenko·
lead-slop /ˈliːd slɑːp/ noun; also attributive computationally generated drug candidates that score well in silico but dissolve into disappointment on contact with the wet-lab
English
0
0
8
184
Kat
Kat@katyenko·
The future we bet on at @muni_bio is one where agentic workflows become the default in early drug discovery. In our latest post, we walk through a pipeline designed and executed autonomously using @RowanSci tools for our TBXT hackathon. These candidates were submitted to @onepot_ai alongside compounds designed by other human participants. As agentic loops continue to reshape every part of the drug discovery pipeline, scientists and engineers need to think about grounding their in-silico loops in the wet lab to avoid lead-slop. muni.bio/research/muni-…
English
3
10
62
6.3K
Kat retweetledi
Dasha
Dasha@dasha_shunina·
people spent their Saturday off... trying to cure Ebola. that's the hackathon @OpenRouter got to support last weekend. this is what "building for good" should look like.
Kat@katyenko

Thank you to everyone who spent their Saturday with us at Ebolathon—and congratulations to the BuilderBlockers, Michael, @cydatio, and Rafal, who took an interesting synthon-aware Thompson Sampling approach to guide compound selection. All submissions have been sent to our co-host, @onepot_ai, for synthesis before moving into pseudovirus assay testing. All assay data will be made public. A special thank you to our judges—@andrei_tyrin, Andrii Kyrylchuk, Amy He, and Nina Jovic—for lending their time and expertise. And thank you to our sponsors—@RowanSci, @boltz_bio, @OpenRouter, @modal, and @anyscalecompute for giving participants access to the models and compute needed to explore the challenge, and to @posthog for hosting us in their wonderful space!

English
0
2
6
635
Kat
Kat@katyenko·
"Bože Pomozi Croatia" ("God Help Croatia" in English) is the richer origin story for BPC-157
Kat tweet media
Română
1
0
5
369
Kat
Kat@katyenko·
@Cas9Bandit thanks so much for joining and asking all the good questions! your team was great
English
0
0
1
96
💦 Dr. Alex Gener (he/él)
💦 Dr. Alex Gener (he/él)@Cas9Bandit·
This was fun! Excited to see how the leads turn out!
Kat@katyenko

Thank you to everyone who spent their Saturday with us at Ebolathon—and congratulations to the BuilderBlockers, Michael, @cydatio, and Rafal, who took an interesting synthon-aware Thompson Sampling approach to guide compound selection. All submissions have been sent to our co-host, @onepot_ai, for synthesis before moving into pseudovirus assay testing. All assay data will be made public. A special thank you to our judges—@andrei_tyrin, Andrii Kyrylchuk, Amy He, and Nina Jovic—for lending their time and expertise. And thank you to our sponsors—@RowanSci, @boltz_bio, @OpenRouter, @modal, and @anyscalecompute for giving participants access to the models and compute needed to explore the challenge, and to @posthog for hosting us in their wonderful space!

English
1
0
5
426
Kat retweetledi
Dom
Dom@dominik_scherm·
To this day, Bundibugyo ebolavirus (BDBV), the species driving the current outbreak in DRC, has zero approved drugs. As part of the Cure Ebola hackathon in SF last Saturday, we trained an RL agent to discover BDBV drugs and found 10 purchasable molecules now heading to a real assay. Here is what we built 🧵
GIF
English
6
13
67
5.7K
Derek Alia
Derek Alia@derekalia·
147.4M tokens during the Ebola hackathon. Grok 4.5 led by a wide margin with 79.8M, more than half of all tokens used. Not surprising. When we were researching the problem internally, Grok gave us the least amount of trouble by far. Congrats to the @xai team for building a model that’s useful across a range of scientific problems. @grok how did you feel about working on these problems?
Derek Alia tweet media
English
1
0
3
146
Derek Alia
Derek Alia@derekalia·
That's a wrap for SF's first biodefense hackathon. Huge thanks to every team that showed up for the Ebola design challenge this weekend. 29 bio/chem tool models. 1,908 runs. One very busy Saturday on muni.bio Really cool to watch how people combined the latest agents, docking, ADMET, library screens, and novel techniques on a real open science problem. Super excited to see what comes back from the wet lab. Special thanks to @andrei_tyrin and @daniil_boiko from @onepot_ai and @CorinWagen from @RowanSci for powering so much of the stack.
Derek Alia tweet mediaDerek Alia tweet mediaDerek Alia tweet mediaDerek Alia tweet media
English
2
5
18
2K
Kat
Kat@katyenko·
Thank you to everyone who spent their Saturday with us at Ebolathon—and congratulations to the BuilderBlockers, Michael, @cydatio, and Rafal, who took an interesting synthon-aware Thompson Sampling approach to guide compound selection. All submissions have been sent to our co-host, @onepot_ai, for synthesis before moving into pseudovirus assay testing. All assay data will be made public. A special thank you to our judges—@andrei_tyrin, Andrii Kyrylchuk, Amy He, and Nina Jovic—for lending their time and expertise. And thank you to our sponsors—@RowanSci, @boltz_bio, @OpenRouter, @modal, and @anyscalecompute for giving participants access to the models and compute needed to explore the challenge, and to @posthog for hosting us in their wonderful space!
Kat tweet mediaKat tweet mediaKat tweet mediaKat tweet media
English
3
8
26
2.4K
Kat
Kat@katyenko·
@sergiomaresd thanks for coming and good to see you again!!
English
1
0
4
66
Kat retweetledi
Sergio Mares @ ICML
Sergio Mares @ ICML@sergiomaresd·
today i went to a hackaton on small molecule design for ebola. If you thought minibidner design was hard holy sh
English
5
3
39
6K
Kat
Kat@katyenko·
Thanks to @posthog for hosting such a crazy ambitious event, and to @OpenRouter and @modal for sponsoring credits. Excited to bring together companies shaping the future of bio x AI: @onepot_ai -- providing synthesis with AI and robots at breakneck speed @RowanSci -- allowing scientists to access the latest molecular design tools @boltz_bio -- building generative models in biology and chemistry @muni_bio -- creating the agentic environments for science We’re moving away from manual coordination into agentic self-learning loops, a shift that positions U.S. biotech to compete on innovation and not labor arbitrage. To help shape this future, come join us on Saturday luma.com/2rbgfn69
Kat tweet media
English
0
8
22
1.3K
Kat retweetledi
Niko McCarty.
Niko McCarty.@NikoMcCarty·
Our @AsimovPress book on the history of the molecular biology lab is nearly finished. It's currently 455 pages, with 100+ images. Some previews below. We've spent the last 6+ months making this book as beautiful, and detailed, as possible. And we're excited for you to hold it.
Niko McCarty. tweet mediaNiko McCarty. tweet mediaNiko McCarty. tweet mediaNiko McCarty. tweet media
English
41
70
617
33.1K
Kat
Kat@katyenko·
The example I give in the post refers to being able to change the score depending on context of your experiment. In the broad term of "therapeutics", you can have agentic loops at different parts of the process: 1. solely for reading literature (i.e. finding a target) 2. generating candidates -> triaging / scoring 3. wet-lab integration to feed back into the model(s) This post mostly describes 2, but loops can be applied in a lot of places/can be orchestrated in different ways. The point is having the agent be able to hypothesize + generate things, send to a model (or lab), receive something in return, make decisions based on those returns, and repeat
English
0
0
0
36
Rafael Rolli
Rafael Rolli@RafaDeSci·
@katyenko What do you mean by agentic loops in therapeutics. Could you give an example?
English
1
0
1
61
Kat
Kat@katyenko·
As agentic loops in therapeutics become all the rage, it’ll be increasingly valuable to think about what score you’re maximizing. ipSAE is specifically good for prediction of positioning and alignment of residues at a protein-protein interface. But what if you care about alternatives like interface quality metrics (pDockQ, LIS), developability properties (hydrophobicity, viscosity, solubility), or immunogenicity/tox? We explored @NVIDIAHealth's Proteina-Complexa’s Composite Reward Model, which lets you customize the attributes and weights in a scoring function to guide agents toward the therapeutic goals that matter most to you. In this research article, we show how adjusting the weights changes the agent’s strategy to allow the model to adapt to different therapeutic design objectives. muni.bio/research/scori…
English
2
5
34
4.9K
Kat
Kat@katyenko·
Excited to have @OpenRouter as a sponsor of our Ebola hackathon this Saturday! Inside @muni_bio, users are able to switch between agents to solve different parts of their scientific problems. This feature wouldn’t be possible without OpenRouter. If you’re interested in science x AI, come join: luma.com/2rbgfn69
Kat tweet media
English
2
1
21
1.4K