Micha Sam Brickman Raredon
@msbr89.bsky.social
Scientist/Engineer. Assistant Professor of Anesthesiology @ Yale. Tissue Biology, Lung Regeneration, Data Visualization. Here to learn. https://RaredonLab.com
December 16, 2025 at 7:49 PM
Our landing point has been to visualize these connectivity fields as three-dimensional surfaces that you can rotate in space!
Here are a few selected fields from a single location in a mouse embryo capturing the developing ribs, lung, and diaphragm:
Interactive HTML: figshare.com/s/2f66def6d2...
Here are a few selected fields from a single location in a mouse embryo capturing the developing ribs, lung, and diaphragm:
Interactive HTML: figshare.com/s/2f66def6d2...
December 15, 2025 at 2:54 PM
Our landing point has been to visualize these connectivity fields as three-dimensional surfaces that you can rotate in space!
Here are a few selected fields from a single location in a mouse embryo capturing the developing ribs, lung, and diaphragm:
Interactive HTML: figshare.com/s/2f66def6d2...
Here are a few selected fields from a single location in a mouse embryo capturing the developing ribs, lung, and diaphragm:
Interactive HTML: figshare.com/s/2f66def6d2...
This lead us down a beautiful rabbit-hole experimenting with the relationship between high-dimensional connectivity and the resolution of spatial tissue domains.
Note that in all of these graphs, we are exclusively studying ligand-receptor information, no intracellular genes:
Note that in all of these graphs, we are exclusively studying ligand-receptor information, no intracellular genes:
December 15, 2025 at 2:48 PM
This lead us down a beautiful rabbit-hole experimenting with the relationship between high-dimensional connectivity and the resolution of spatial tissue domains.
Note that in all of these graphs, we are exclusively studying ligand-receptor information, no intracellular genes:
Note that in all of these graphs, we are exclusively studying ligand-receptor information, no intracellular genes:
To evaluate, we needed to quantify the high-dimensional structure of these 'connectivity fields'. So, we performed unsupervised clustering and measured the number and the spatial entropy (coherence) of the resulting groups. We found that a stoichiometrically-informed convolution maximized structure:
December 15, 2025 at 2:44 PM
To evaluate, we needed to quantify the high-dimensional structure of these 'connectivity fields'. So, we performed unsupervised clustering and measured the number and the spatial entropy (coherence) of the resulting groups. We found that a stoichiometrically-informed convolution maximized structure:
But this modeling was done on the ligand and the receptor individually. We wanted to predict density fields for ligand-receptor binding! We therefore experimented with ways of convolving these outputs, comparing established methods with next-gen ideas informed by ligand-receptor binding kinetics:
December 15, 2025 at 2:32 PM
But this modeling was done on the ligand and the receptor individually. We wanted to predict density fields for ligand-receptor binding! We therefore experimented with ways of convolving these outputs, comparing established methods with next-gen ideas informed by ligand-receptor binding kinetics:
But how to approach? There are many well-established spatial statistical models for data with exactly this form. We tested and compared three in particular: a Poisson, Zero-Inflated Poisson (ZIP), and Zero-Adjusted Poisson (ZAP), and found that the Poisson model was the top performing:
December 15, 2025 at 2:25 PM
But how to approach? There are many well-established spatial statistical models for data with exactly this form. We tested and compared three in particular: a Poisson, Zero-Inflated Poisson (ZIP), and Zero-Adjusted Poisson (ZAP), and found that the Poisson model was the top performing:
For one, we realized almost immediately that we needed to think carefully about what data to use as input.
Most single-cell and spatial pipelines use count-normalized data. But we learned that this was likely not the best starting point for the tissue-scale modeling we had in mind.
Most single-cell and spatial pipelines use count-normalized data. But we learned that this was likely not the best starting point for the tissue-scale modeling we had in mind.
December 15, 2025 at 2:09 PM
For one, we realized almost immediately that we needed to think carefully about what data to use as input.
Most single-cell and spatial pipelines use count-normalized data. But we learned that this was likely not the best starting point for the tissue-scale modeling we had in mind.
Most single-cell and spatial pipelines use count-normalized data. But we learned that this was likely not the best starting point for the tissue-scale modeling we had in mind.
If handled thoughtfully, we realized that these count values could be used to infer continuous fields estimating local ligand density, local receptor density, and the local probability of ligand-receptor binding events. We had a loose plan, but many questions about specifics...
December 15, 2025 at 2:04 PM
If handled thoughtfully, we realized that these count values could be used to infer continuous fields estimating local ligand density, local receptor density, and the local probability of ligand-receptor binding events. We had a loose plan, but many questions about specifics...
We found ourselves particularly interested in modeling local ligand-receptor interaction probability densities. This is because these data have a wealth of information regarding omics-level (high-dimensional) ligand and receptor transcription, in space:
December 15, 2025 at 2:00 PM
We found ourselves particularly interested in modeling local ligand-receptor interaction probability densities. This is because these data have a wealth of information regarding omics-level (high-dimensional) ligand and receptor transcription, in space:
There are many things that we can think about measuring and modeling spatially using these types of data. Aaron and I have now had years of conversations about the different ways researchers might approach these data and the pros/cons of applying different perspectives.
December 15, 2025 at 1:58 PM
There are many things that we can think about measuring and modeling spatially using these types of data. Aaron and I have now had years of conversations about the different ways researchers might approach these data and the pros/cons of applying different perspectives.
We were interested in seeing if we could use spatial statistics, well developed in geospatial statistics, to study morphogenic interactions in living tissues.
Spatial-omics methods are now producing rich datasets across entire tissues and organisms, allowing tissue-scale analysis:
Spatial-omics methods are now producing rich datasets across entire tissues and organisms, allowing tissue-scale analysis:
December 15, 2025 at 1:56 PM
We were interested in seeing if we could use spatial statistics, well developed in geospatial statistics, to study morphogenic interactions in living tissues.
Spatial-omics methods are now producing rich datasets across entire tissues and organisms, allowing tissue-scale analysis:
Spatial-omics methods are now producing rich datasets across entire tissues and organisms, allowing tissue-scale analysis:
In the body, there are many forms of cellular input and output. Cells sense their environments, process this information in a state-dependent way, and then respond accordingly. These input/output loops are major drivers of tissue development and regulators of collective cellular behavior.
December 15, 2025 at 1:52 PM
In the body, there are many forms of cellular input and output. Cells sense their environments, process this information in a state-dependent way, and then respond accordingly. These input/output loops are major drivers of tissue development and regulators of collective cellular behavior.
20 years ago, Aaron Osgood-Zimmerman and I would entertain ourselves asking questions like, "How many crickets are in field?" or "How many cells are in the body?"
Aaron became a statistician, and I become a biologist.
We have now collaborated on an interesting exploration: doi.org/10.64898/202...
Aaron became a statistician, and I become a biologist.
We have now collaborated on an interesting exploration: doi.org/10.64898/202...
December 15, 2025 at 1:49 PM
20 years ago, Aaron Osgood-Zimmerman and I would entertain ourselves asking questions like, "How many crickets are in field?" or "How many cells are in the body?"
Aaron became a statistician, and I become a biologist.
We have now collaborated on an interesting exploration: doi.org/10.64898/202...
Aaron became a statistician, and I become a biologist.
We have now collaborated on an interesting exploration: doi.org/10.64898/202...
December 3, 2025 at 12:56 PM
October 1, 2025 at 1:21 AM
"Single cell RNA seq reveals the pro-regenerative phenotype of thrombospondin-2 deficient dermal fibroblasts"
www.nature.com/articles/s41...
Amazing work from Huang et al!
www.nature.com/articles/s41...
Amazing work from Huang et al!
September 2, 2025 at 12:46 PM
"Single cell RNA seq reveals the pro-regenerative phenotype of thrombospondin-2 deficient dermal fibroblasts"
www.nature.com/articles/s41...
Amazing work from Huang et al!
www.nature.com/articles/s41...
Amazing work from Huang et al!
'Single cell decomposition of multicellular aging programs associated with impaired lung regeneration'
www.biorxiv.org/content/10.1...
www.biorxiv.org/content/10.1...
August 22, 2025 at 1:07 PM
'Single cell decomposition of multicellular aging programs associated with impaired lung regeneration'
www.biorxiv.org/content/10.1...
www.biorxiv.org/content/10.1...
'p53 maintains lineage fidelity during lung capillary injury-repair in neonatal hyperoxia'
insight.jci.org/articles/vie...
insight.jci.org/articles/vie...
August 22, 2025 at 1:06 PM
'p53 maintains lineage fidelity during lung capillary injury-repair in neonatal hyperoxia'
insight.jci.org/articles/vie...
insight.jci.org/articles/vie...
August 2, 2025 at 5:31 PM
The Edge of Objectivity, page 53:
August 2, 2025 at 12:11 PM
The Edge of Objectivity, page 53:
"The interplay between biomechanics and cell kinetics explains the spatial pattern in liver fibrosis"
www.biorxiv.org/content/10.1...
www.biorxiv.org/content/10.1...
August 2, 2025 at 11:45 AM
"The interplay between biomechanics and cell kinetics explains the spatial pattern in liver fibrosis"
www.biorxiv.org/content/10.1...
www.biorxiv.org/content/10.1...
"thinking for yourself is a frictional activity not a statistical correlation. An AI-mediated essay plan has already missed the point by bypassing the student's own capacity to develop and substantiate propositions about the world."
from @danmcquillan.bsky.social
danmcquillan.org/cpct_seminar...
from @danmcquillan.bsky.social
danmcquillan.org/cpct_seminar...
June 23, 2025 at 12:37 PM
"thinking for yourself is a frictional activity not a statistical correlation. An AI-mediated essay plan has already missed the point by bypassing the student's own capacity to develop and substantiate propositions about the world."
from @danmcquillan.bsky.social
danmcquillan.org/cpct_seminar...
from @danmcquillan.bsky.social
danmcquillan.org/cpct_seminar...
OMG I am so glad someone finally did this.
Thank you 🙏 @lelandmcinnes.bsky.social
This will now consume hours and hours of my time.
lmcinnes.github.io/datamapplot_...
Thank you 🙏 @lelandmcinnes.bsky.social
This will now consume hours and hours of my time.
lmcinnes.github.io/datamapplot_...
June 23, 2025 at 12:12 PM
OMG I am so glad someone finally did this.
Thank you 🙏 @lelandmcinnes.bsky.social
This will now consume hours and hours of my time.
lmcinnes.github.io/datamapplot_...
Thank you 🙏 @lelandmcinnes.bsky.social
This will now consume hours and hours of my time.
lmcinnes.github.io/datamapplot_...
June 15, 2025 at 10:29 PM