Ian Quigley @allmeasures

Cofounder, @LeashBio. Ex-Recursion, Arima Genomics, Nanocellect, Salk, UT Austin, Baylor College of Medicine, Rice. He/him. Salt Lake City, UT Joined May 2010

Tweets

7K
Followers

1K
Following

3K
Likes

20K

Samuel Hume @DrSamuelBHume

14 hours ago

For the first time, most biotech deal value now goes to assets from China:

3 11 55 11K 21

View Details

No scaling laws for single-cell foundation models: when bigger atlases stop teaching the model anything In language and vision, the recipe has been simple: more data, bigger models, better performance. Single-cell biology borrowed that playbook. Foundation models for transcriptomics jumped from 1 million cells to atlases of over 100 million, on the assumption that scale would unlock the same gains. Alan DenAdel and coauthors put that assumption to the test, and the result is sobering. Working from a 22.2-million-cell corpus, they pretrained 400 models across five architectures (from PCA and a variational autoencoder up to the Geneformer transformer) and ran 6,400 evaluation experiments. They varied not just dataset size (1% to 75%) but also diversity, using cell-type re-weighting and geometric sketching to deliberately enrich rare cell types and transcriptional states. The finding: performance saturates almost immediately. On cell-type classification, batch integration, and perturbation prediction, most models hit their ceiling at roughly 1% of the corpus, about 200,000 cells. Beyond that, adding millions more cells changed essentially nothing. More diversity didn't help. Even spiking in genome-scale Perturb-seq data, to give the models perturbed phenotypes rather than just healthy ones, failed to move the needle. Larger models did score better overall, but they too plateaued early on data. Two points stood out. Simple baselines (PCA, logistic regression) often matched or beat the transformers. And the strongest model, SCimilarity, won not because of size but because its contrastive training objective is aligned with the downstream task. For single-cell data, what you train on and how you frame the objective matters far more than how much you collect. This reframes a quiet but expensive habit. In drug discovery, biotech, and any pipeline leaning on cell atlases, the instinct to keep scaling pretraining corpora may be burning compute for no return. The real leverage sits elsewhere: curating high-quality, task-relevant data and matching the training objective to the actual question you're trying to answer. Paper: DenAdel et al., journal license | doi.org/10.1038/s41592…

13 81 346 78K 248

View Details

Ron Alfa @Ronalfa

a day ago

its a very small community in ai bio, everyone knows each other. we are working on human health, integrity is a big deal.

11 8 173 12K 23

View Details

Ian Quigley @allmeasures

3 days ago

@JeremyNguyenPhD onlinelibrary.wiley.com/doi/10.1111/st…

0 0 0 27 0

View Details

Ian Quigley @allmeasures

7 days ago

I'm on a mission to make this plot as publicly recognizable as the one with the holes in the WW2 airplane

Ian Quigley @allmeasures

a week ago

8 80 1K 639K 387

35 140 4K 442K 790

View Details

Thea Don-Siemion @StationryBandit

6 days ago

Done.

Ian Quigley @allmeasures

7 days ago

I'm on a mission to make this plot as publicly recognizable as the one with the holes in the WW2 airplane

35 140 4K 442K 790

0 1 13 816 0

View Details

Anshul Kundaje @anshulkundaje

7 days ago

Big funding agencies / VCs / companies etc interested in powering true breakthroughs in AI for biology should fund tech dev efforts that can produce & scale the right types of data to push models out of zones they struggle in. 8/

1 1 16 1K 4

View Details

Ian Quigley @allmeasures

7 days ago

I like this but the Takifugu rubripes genome is ~1/8 as big as ours, has about the same number of genes, has bones gut neurons etc, maybe start there?

Sudarshan Pinglay @sudpinglay

a week ago

How much of the human genome is essential? Two pieces out today from our lab: 1) a method to map essential genomic intervals at gigabase scale, and 2) an argument that it's time to consider synthesizing a minimal human genome. biorxiv.org/content/10.648… nature.com/articles/d4158…

8 47 178 40K 106

0 0 1 332 0

View Details

Ian Quigley @allmeasures

a week ago

Oded Rechavi @OdedRechavi

a week ago

Most experiments fail, and negative results rarely get published. This means LLMs are unaware of the outcomes of most experiments.

151 745 8K 574K 713

8 80 1K 639K 387

View Details

Ian Quigley @allmeasures

a week ago

@bauer_lesavage @_DimensionCap Thanks for the Clever Hans shoutout!

0 0 4 89 0

View Details

Ron Alfa @Ronalfa

a week ago

Another great research piece by @_DimensionCap @bauer_lesavage on Training Data for Bio AI. Models will only be as good as the underlying data, and the biology they learn will be constrained by the limitations of that data. We need to think deeply about scaling the best quality data to solve human bio with AI. Our thesis from day 1 at @NOETIK_ai.

Bauer LeSavage @bauer_lesavage

a week ago

The training data market has exploded for LLMs and bio foundation models are next. But biological data is extremely complex and requires a data generation playbook that prioritizes quality over immediate scale. @_DimensionCap Research article live now! research.dimensioncap.com/p/on-training-…

9 15 106 25K 90

6 13 84 12K 43

View Details

Ian Quigley @allmeasures

a week ago

Probably noisy as hell but if you were smart about the chemical library design .. so crazy it just might work

0 0 1 96 1

View Details

Ian Quigley @allmeasures

a week ago

Somebody needs to do this and then throw a DEL at it and do in situ sequencing of the DEL barcodes. You could get cellular chemistry tropism at scale

Ali H. Shaib @AliHShaib

a week ago

Hello world, meet 1,000× Expansion Microscopy. 1,000,000,000× expansion by volume! A gel that starts at a few centimeters will then expand to the volume of an Olympic swimming pool. biorxiv.org/content/10.648… In our new bioRxiv preprint, work carried out between MIT and UMG, led by

12 140 597 96K 319

2 3 19 3K 4

View Details

Ian Quigley @allmeasures

a week ago

This is such a banger of an ASCO

Avi Roy @agingroy

a week ago

A new cancer drug works like a molecular handcuff. One arm grabs a passing immune cell. The other grabs the tumor. It clamps them together and forces the kill, whether the cancer agrees to the meeting or not. That trick cleared blood cancers a decade ago. Blinatumomab still does

4 71 376 33K 103

2 0 3 1K 2

View Details

Tasuku Ishida @TasukuIshida

2 weeks ago

Animals have expanded the evolutionary legacy of unicellular ancestors in blood cells pnas.org/doi/10.1073/pn…

0 2 4 426 1

View Details

IdoAmitLab @IdoAmitLab

2 weeks ago

For those of you who think macrophages do almost everything, rethink “almost.” They are programmable cells, recruited into whatever new function evolution needs. science.org/doi/10.1126/sc…

4 68 290 72K 139

View Details

Romain Lopez @_romain_lopez_

2 weeks ago

We built a joint experimental and computational platform for scalable multi-modal single-cell chemical screens — profiling RNA, protein (including phospho-signaling), and chromatin accessibility responses to thousands of small molecule perturbations in parallel. biorxiv.org/content/10.648…

2 40 180 14K 120

View Details

owl @owl_posting

2 weeks ago

Avi Roy @agingroy

a week ago

4 71 376 33K 103

10 45 679 43K 87

View Details

Ian Quigley @allmeasures

3 weeks ago

So let's help where we can now, and invent ways to measure things where we can't. A mechanistic model of a cell is probably going to be pretty bad right now, since we can't possibly measure all of the important aspects. (n/n)

1 0 1 45 0

View Details

Ian Quigley @allmeasures

3 weeks ago

Being thoughtful about what we can model from existing measurement modalities is going to be how we help patients in the near term, and we should also be thoughtful about what technologies to build to measure things we can't model yet.

1 0 1 48 0

View Details

Ian Quigley @allmeasures

3 weeks ago

Physical dynamics of proteins in an intact tissue measured over time, driven by cytoskeletal interactions. Biology like this is why virtual cell efforts are woefully incomplete. Nobody is able to detect this kind of stuff at scale! (1/n)

kuba sędziński @ksedzinski

3 weeks ago

💥 🚀 New preprint! 🎉🥳 How do hundreds of organelles organize themselves into near-perfect patterns inside a cell, without a blueprint? We dive deep into how basal bodies (BBs) self-organize in MCCs - and how actin actively tunes their dynamics into order 🍪 🧵👇 (1/17)