3/
Big thanks to @denisschapiro.bsky.social , @miguelib.bsky.social , @kbestak.bsky.social , and @tanevski.bsky.social - and to everyone else who helped along the way 🤝. This project took its time, and I couldn’t have brought it this far without you.
2/
🫵 If you want to try out COZI, it is available as a package in Python (pypi.org/project/cozi...), R (github.com/SchapiroLabo...) and now also as a part of IMCRtools (github.com/BodenmillerG...)!
1/
A PhD milestone 👏 My work on comparing neighbor preference methods for spatial data analysis is out in Nature Communications.
We compared 9+ neighbor preference methods and propose a new approach that combines the most relevant analysis features: COZI.
Read more: www.nature.com/articles/s41...
Never thought I'd co-first author a paper that appears on the screens on campus, that makes me very happy😃
www.cell.com/cell/fulltex...
Check out the paper and get amazing summaries from the EMBL communications team and a thread by @chanyeong-kim.bsky.social
I am excited to share my first work in @schapirolab.bsky.social on Multiple Myeloma (MM) in collaboration with the Standal Lab from NTNU Norway. We use Imaging Mass Cytometry (IMC) on bone marrow biopsies from MM and precursor patients (details below): www.biorxiv.org/content/10.1...
If you missed @chiaraschiller.bsky.social’s talk at #EMBLSpatialBiology you can catch her at Poster 188 this evening
6/ Thank you
This work was led by @chiaraschiller.bsky.social w help from @kbestak.bsky.social and supervision from @miguelib.bsky.social, @tanevski.bsky.social and @denisschapiro.bsky.social
5/ Real data validation
In myocardial infarction tissues, COZI uniquely detects:
- Early neutrophil infiltration
- Monocyte targeting of stressed cardiomyocytes
- Spatial infiltration gradients
4/ One step in the right direction: COZI
COZI (Conditional Z-score) combines advantageous method features by normalizing neighbor counts based on context and providing a z-score. It captures directional preferences and performs robustly across conditions, cell types, and neighborhood structures.
3/ The Comparison
Using simulated tissues with known patterns, we show that existing methods often:
- Miss directionality of NEP
- Struggle with low-abundance cell types
- Lack sensitivity to detect subtle NEP changes
2/ The problem
Cell-cell spatial relationships are key to tissue function, but comparing NEP across datasets or conditions is tricky. Existing methods differ in how they define neighborhoods, count neighbors, and score NEP—leading to inconsistent results. We identified the common building blocks.
Schematic overview of NEP analysis steps (Neighborhood definition, Quantification and NEP score) and the systematic method performance comparison using simulated data for cohort distinction.
1/
Ever wondered how to best quantify cell-cell neighbor preferences in tissues?
We compared 9+ neighbor preference (NEP) methods for analysing spatial omics data and propose a novel approach that combines the most relevant analysis features which we call COZI 🔬✨
Read more: doi.org/10.1101/2025...
We are searching for a highly motivated and talented computational postdoc with an interest in translational spatial omics to join our team in Heidelberg. Apply now!
karriere.klinikum.uni-heidelberg.de/index.php?ac...
A presenter showing a slide on the importance of learning resources for beginners on spatial omics analysis.
screenshot of the new spatialdata-io command line interface for converting common spatial omics technologies to the SpatialData Zarr format.
Quentin Blampey, Lotte Pollaris, @clarencemah.me Laurens Lehner @chiaraschiller.bsky.social @miguelib.bsky.social
We worked on easing the learning curve of the SpatialData framework by improving the documentation and APIs. We prepared new beginner-friendly notebooks and introduced a new more...
@chiaraschiller.bsky.social - Poster 149