Brian DePasquale
@briandepasquale.bsky.social
science. https://briandepasquale.github.io
Reposted by Brian DePasquale
The Batman Effect, just published in NATURE:
When an experimenter dressed as Batman boarded a train, passengers were significantly more likely to offer their seats to another “pregnant” experimenter than if Batman wasn’t present (67.21% vs. 37.66%) #SciComm 1/2
www.nature.com/articles/s44...
When an experimenter dressed as Batman boarded a train, passengers were significantly more likely to offer their seats to another “pregnant” experimenter than if Batman wasn’t present (67.21% vs. 37.66%) #SciComm 1/2
www.nature.com/articles/s44...
Unexpected events and prosocial behavior: the Batman effect - npj Mental Health Research
npj Mental Health Research - Unexpected events and prosocial behavior: the Batman effect
www.nature.com
November 22, 2025 at 12:37 PM
The Batman Effect, just published in NATURE:
When an experimenter dressed as Batman boarded a train, passengers were significantly more likely to offer their seats to another “pregnant” experimenter than if Batman wasn’t present (67.21% vs. 37.66%) #SciComm 1/2
www.nature.com/articles/s44...
When an experimenter dressed as Batman boarded a train, passengers were significantly more likely to offer their seats to another “pregnant” experimenter than if Batman wasn’t present (67.21% vs. 37.66%) #SciComm 1/2
www.nature.com/articles/s44...
🔔 NEW PREPRINT FROM THE LAB ‼️
We introduce a new ML model, LoRAX, for predicting olfactory responses from chemical features, a tricky problem that benefits from progress in ML for biochem. We combine LoRA fine-tuning with protein and chemical foundation models, www.biorxiv.org/content/10.1...
We introduce a new ML model, LoRAX, for predicting olfactory responses from chemical features, a tricky problem that benefits from progress in ML for biochem. We combine LoRA fine-tuning with protein and chemical foundation models, www.biorxiv.org/content/10.1...
Low rank adaptation of chemical foundation models generate effective odorant representations
Featurizing odorants to enable robust prediction of their properties is difficult due to the complex activation patterns that odorants evoke in the olfactory system. Structurally similar odorants can ...
www.biorxiv.org
November 20, 2025 at 1:38 PM
🔔 NEW PREPRINT FROM THE LAB ‼️
We introduce a new ML model, LoRAX, for predicting olfactory responses from chemical features, a tricky problem that benefits from progress in ML for biochem. We combine LoRA fine-tuning with protein and chemical foundation models, www.biorxiv.org/content/10.1...
We introduce a new ML model, LoRAX, for predicting olfactory responses from chemical features, a tricky problem that benefits from progress in ML for biochem. We combine LoRA fine-tuning with protein and chemical foundation models, www.biorxiv.org/content/10.1...
Reposted by Brian DePasquale
Today, we're announcing Kosmos, our newest AI Scientist, available today. Kosmos makes fully autonomous scientific discoveries at scale by analyzing datasets and literature, and is the most powerful agent for science so far. Beta users estimate that Kosmos does 6 months of work in a single day.
November 5, 2025 at 8:10 PM
Today, we're announcing Kosmos, our newest AI Scientist, available today. Kosmos makes fully autonomous scientific discoveries at scale by analyzing datasets and literature, and is the most powerful agent for science so far. Beta users estimate that Kosmos does 6 months of work in a single day.
Reposted by Brian DePasquale
1. New preprint resolving a conundrum in systems neuroscience with an AI scientist, and humans Reilly Tilbury, Dabin Kwon, @haydari.bsky.social, @jacobmratliff.bsky.social, @bio-emergent.bsky.social, @carandinilab.net, @kevinjmiller.bsky.social, @neurokim.bsky.social
www.biorxiv.org/content/10.1...
www.biorxiv.org/content/10.1...
Characterizing neuronal population geometry with AI equation discovery
The visual cortex contains millions of neurons, whose combined activity forms a population code representing visual stimuli. There is, however, a discrepancy between our understanding of this code at ...
www.biorxiv.org
November 14, 2025 at 6:07 PM
1. New preprint resolving a conundrum in systems neuroscience with an AI scientist, and humans Reilly Tilbury, Dabin Kwon, @haydari.bsky.social, @jacobmratliff.bsky.social, @bio-emergent.bsky.social, @carandinilab.net, @kevinjmiller.bsky.social, @neurokim.bsky.social
www.biorxiv.org/content/10.1...
www.biorxiv.org/content/10.1...
First published paper from the depaqlab, led by the supernatural @ryguy.io! Glad to add a new software resource to the neuro community for fitting SSMs, including hierarchical models and switching GLMs, and many others! If you love Julia and SSMs, this is for you!
🚨🚨🚨NEW PAPER ALERT🚨🚨🚨
We present StateSpaceDynamics.jl, an open-source Julia package for fitting probabilistic state-space models (SSMs)—including non-conjugate models like Poisson LDS and switching variants (SLDS). Python has had existing packages that have given this functionality...
We present StateSpaceDynamics.jl, an open-source Julia package for fitting probabilistic state-space models (SSMs)—including non-conjugate models like Poisson LDS and switching variants (SLDS). Python has had existing packages that have given this functionality...
Just published in JOSS: 'StateSpaceDynamics.jl: A Julia package for probabilistic state space models (SSMs)' https://doi.org/10.21105/joss.08077
November 15, 2025 at 2:26 AM
First published paper from the depaqlab, led by the supernatural @ryguy.io! Glad to add a new software resource to the neuro community for fitting SSMs, including hierarchical models and switching GLMs, and many others! If you love Julia and SSMs, this is for you!
Reposted by Brian DePasquale
“The bad review will come from your list of suggested reviewers”
October 31, 2025 at 7:39 AM
“The bad review will come from your list of suggested reviewers”
Reposted by Brian DePasquale
What if we did a single run and declared victory
October 23, 2025 at 2:28 AM
What if we did a single run and declared victory
Reposted by Brian DePasquale
Ouch. This one cuts to the bone.
October 18, 2025 at 8:41 PM
Ouch. This one cuts to the bone.
Reposted by Brian DePasquale
Chemists call for ban on generative AI for chemical structures: "serious errors could damag[e] the next generation of scientists"
As ever, the problem seems to be not AI but lazy application of it and lack of checks. Whether _that_ problem is solvable...
www.chemistryworld.com/news/the-che...
As ever, the problem seems to be not AI but lazy application of it and lack of checks. Whether _that_ problem is solvable...
www.chemistryworld.com/news/the-che...
The chemistry community should ban drawing chemical structures with generative AI, chemists warn
AIs like Microsoft's Copilot, Google's Gemini and OpenAI's ChatGPT still make serious errors rendering structural formulae
www.chemistryworld.com
October 16, 2025 at 12:10 PM
Chemists call for ban on generative AI for chemical structures: "serious errors could damag[e] the next generation of scientists"
As ever, the problem seems to be not AI but lazy application of it and lack of checks. Whether _that_ problem is solvable...
www.chemistryworld.com/news/the-che...
As ever, the problem seems to be not AI but lazy application of it and lack of checks. Whether _that_ problem is solvable...
www.chemistryworld.com/news/the-che...
Reposted by Brian DePasquale
Reposted by Brian DePasquale
"the quickest way to get a collaborator to return edits on your manuscript is to write down their affiliation wrong" (ancient academic proverb)
October 6, 2025 at 3:21 AM
"the quickest way to get a collaborator to return edits on your manuscript is to write down their affiliation wrong" (ancient academic proverb)
Reposted by Brian DePasquale
Can one bring together Reinforcement learning and Drift Diffusion models to understand collective foraging ?
Congrads to Jonathan Marienhagen , Lisa Blum Moyse and Dominik Deffner on this new study. Very happy that I was part of this collaboration.
Preprint here: osf.io/preprints/ps...
Congrads to Jonathan Marienhagen , Lisa Blum Moyse and Dominik Deffner on this new study. Very happy that I was part of this collaboration.
Preprint here: osf.io/preprints/ps...
September 16, 2025 at 10:14 AM
Can one bring together Reinforcement learning and Drift Diffusion models to understand collective foraging ?
Congrads to Jonathan Marienhagen , Lisa Blum Moyse and Dominik Deffner on this new study. Very happy that I was part of this collaboration.
Preprint here: osf.io/preprints/ps...
Congrads to Jonathan Marienhagen , Lisa Blum Moyse and Dominik Deffner on this new study. Very happy that I was part of this collaboration.
Preprint here: osf.io/preprints/ps...
Reposted by Brian DePasquale
PLOS Comput. Biol.: Stochastic activity in low-rank recurrent neural networks
https://journals.plos.org/ploscompbiol/article?id=10.1371/journal.pcbi.1013371
https://journals.plos.org/ploscompbiol/article?id=10.1371/journal.pcbi.1013371
August 20, 2025 at 1:03 PM
PLOS Comput. Biol.: Stochastic activity in low-rank recurrent neural networks
https://journals.plos.org/ploscompbiol/article?id=10.1371/journal.pcbi.1013371
https://journals.plos.org/ploscompbiol/article?id=10.1371/journal.pcbi.1013371
Reposted by Brian DePasquale
“NeuroAI should not remain limited to learning statistical relationships, but should also help in building mechanistic and causal models of neural activity. These models will incorporate biological properties of neural circuits, including cellular characteristics and network properties.”
💯💯💯
💯💯💯
We were asked to write a commentary on NeuroAI for Nat. Rev. Neuro. with Sadra Sadeh, so here it is: www.nature.com/articles/s41...
www.nature.com
August 8, 2025 at 12:43 PM
“NeuroAI should not remain limited to learning statistical relationships, but should also help in building mechanistic and causal models of neural activity. These models will incorporate biological properties of neural circuits, including cellular characteristics and network properties.”
💯💯💯
💯💯💯
Reposted by Brian DePasquale
New preprint from @darbly.bsky.social @briandepasquale.bsky.social and my labs! If you love (or hate) mosquitoes, have a look:
doi.org/10.1101/2025...
We used TEM to describe the circuitry used for CO2 detection by mosquitoes. As usual with mosquitoes, nothing is as expected! 🌬️🦟
doi.org/10.1101/2025...
We used TEM to describe the circuitry used for CO2 detection by mosquitoes. As usual with mosquitoes, nothing is as expected! 🌬️🦟
Recurrent connectivity supports carbon dioxide sensitivity in Aedes aegypti mosquitoes
The mosquito Aedes aegypti′s human host-seeking behavior depends on the integration of multiple sensory cues. One of these cues, carbon dioxide (CO2), gates odorant and heat pathways and activates hos...
doi.org
July 31, 2025 at 12:28 PM
New preprint from @darbly.bsky.social @briandepasquale.bsky.social and my labs! If you love (or hate) mosquitoes, have a look:
doi.org/10.1101/2025...
We used TEM to describe the circuitry used for CO2 detection by mosquitoes. As usual with mosquitoes, nothing is as expected! 🌬️🦟
doi.org/10.1101/2025...
We used TEM to describe the circuitry used for CO2 detection by mosquitoes. As usual with mosquitoes, nothing is as expected! 🌬️🦟
Reposted by Brian DePasquale
June 19, 2025 at 4:31 PM
Reposted by Brian DePasquale
How do brain circuits evolve? We started looking for some answers by using synapse-resolution cross-species comparative connectomics on an entire olfactory circuit 👇
bit.ly/44aVm9E
bit.ly/44aVm9E
June 12, 2025 at 11:04 AM
How do brain circuits evolve? We started looking for some answers by using synapse-resolution cross-species comparative connectomics on an entire olfactory circuit 👇
bit.ly/44aVm9E
bit.ly/44aVm9E
Reposted by Brian DePasquale
Can #AI become a true scientist? @ninamiolane.bsky.social explores how new technologies are reshaping scientific discovery, and why human expertise remains essential as we enter a new era of research powered by intelligent algorithms. 🧪
plos.io/4kZi0aB
plos.io/4kZi0aB
The fifth era of science: Artificial scientific intelligence
Can AI become a true scientist? This Perspective explores how new technologies are reshaping scientific discovery, and why human expertise remains essential as we enter a new era of research powered b...
plos.io
June 11, 2025 at 11:46 AM
Can #AI become a true scientist? @ninamiolane.bsky.social explores how new technologies are reshaping scientific discovery, and why human expertise remains essential as we enter a new era of research powered by intelligent algorithms. 🧪
plos.io/4kZi0aB
plos.io/4kZi0aB
Tired of your boring old DDM? Check out recent work by grad student @ryguy.io that introduces a *state-dependent* DDM using an underlying HMM. We find mice change their speed-accuracy strategy from trial to trial! Code provided! (in Julia of course!)
🚨PREPRINT ALERT🚨
"A State–Dependent Drift Diffusion Model Reveals Mice Actively Explore A Speed–Accuracy Continuum During Decision-Making" is now available on biorxiv! Check it out below 👇
www.biorxiv.org/content/10.1...
In this work we find that mice are not just static mindless decision makers.
"A State–Dependent Drift Diffusion Model Reveals Mice Actively Explore A Speed–Accuracy Continuum During Decision-Making" is now available on biorxiv! Check it out below 👇
www.biorxiv.org/content/10.1...
In this work we find that mice are not just static mindless decision makers.
A State-Dependent Drift Diffusion Model Reveals Mice Actively Explore A Speed–Accuracy Continuum During Decision-Making
Understanding how animals shift between different decision-making strategies is critical for bridging normative models with naturalistic behavior. While drift diffusion models (DDMs) provide a powerfu...
www.biorxiv.org
June 10, 2025 at 6:48 PM
Tired of your boring old DDM? Check out recent work by grad student @ryguy.io that introduces a *state-dependent* DDM using an underlying HMM. We find mice change their speed-accuracy strategy from trial to trial! Code provided! (in Julia of course!)
Reposted by Brian DePasquale
90% of doing science is being open to new ideas.
May 31, 2025 at 6:14 PM
90% of doing science is being open to new ideas.
Reposted by Brian DePasquale
Unsurprising, but nevertheless tragic. Makes it very hard for integrators and multidisciplinarians to exist in academia. #academicsky
The further a researcher or inventor moves from their previous work, the less cited their latest work will be
https://go.nature.com/44VjEW9
https://go.nature.com/44VjEW9
Shifting research focus comes with the risk of reduced impact
An analysis of millions of scientific papers and patents shows that the further a researcher or inventor moves from their previous work, the less cited their latest work will be.
go.nature.com
June 1, 2025 at 2:43 AM
Unsurprising, but nevertheless tragic. Makes it very hard for integrators and multidisciplinarians to exist in academia. #academicsky
Reposted by Brian DePasquale
Hivemind, what's the best approach/tool to aggregate neural data across sessions (and animals) in dimensionality reduction analyses, particularly when it's not known how overlapping the pool of neurons are over different sessions?
May 10, 2025 at 3:30 PM
Hivemind, what's the best approach/tool to aggregate neural data across sessions (and animals) in dimensionality reduction analyses, particularly when it's not known how overlapping the pool of neurons are over different sessions?
Reposted by Brian DePasquale
"The core strategy is to slow NIH spending to generate surpluses...then use surpluses as justification for slashing the budget...unspent [$] won’t be because NIH doesn’t have good science to spend on [but] because the administration has blocked spending" donmoynihan.substack.com/p/the-nih-bu...
donmoynihan.substack.com
April 29, 2025 at 5:53 PM
"The core strategy is to slow NIH spending to generate surpluses...then use surpluses as justification for slashing the budget...unspent [$] won’t be because NIH doesn’t have good science to spend on [but] because the administration has blocked spending" donmoynihan.substack.com/p/the-nih-bu...
Reposted by Brian DePasquale
Using pretrained vision-language models to align neural signals from different modalities:
arxiv.org/pdf/2407.14020
Cool stuff, though I would still love to see if we could align different neural modalities without the pretrained VLMs...
🧠📈 🧪 #NeuroAI
arxiv.org/pdf/2407.14020
Cool stuff, though I would still love to see if we could align different neural modalities without the pretrained VLMs...
🧠📈 🧪 #NeuroAI
arxiv.org
April 29, 2025 at 6:40 PM
Using pretrained vision-language models to align neural signals from different modalities:
arxiv.org/pdf/2407.14020
Cool stuff, though I would still love to see if we could align different neural modalities without the pretrained VLMs...
🧠📈 🧪 #NeuroAI
arxiv.org/pdf/2407.14020
Cool stuff, though I would still love to see if we could align different neural modalities without the pretrained VLMs...
🧠📈 🧪 #NeuroAI