iiinova — where your digital self keeps evolving

Pablo G. Camara @camarapg.bsky.social · 12/06/2026

We are really grateful to @chanzuckerberg.bsky.social and the NIH BRAIN Initiative for supporting this effort.

Pablo G. Camara @camarapg.bsky.social · 12/06/2026

We’d love to hear from people who try it, especially on datasets where cell morphology has made localization analysis difficult.

110

Pablo G. Camara @camarapg.bsky.social · 12/06/2026

CellAligner is included in the new release of CAJAL, our open-source Pyhton package for the analysis of cell morphology and subcellular organization: github.com/CamaraLab/CA...

github.com

GitHub - CamaraLab/CAJAL: A Python package using Gromov-Wasserstein distance to compare cell shapes

A Python package using Gromov-Wasserstein distance to compare cell shapes - CamaraLab/CAJAL

100

Pablo G. Camara @camarapg.bsky.social · 12/06/2026

We also developed a scalable version of the approach for datasets containing millions of cells by distilling CellAligner into a deep metric learning network.

100

Pablo G. Camara @camarapg.bsky.social · 12/06/2026

In neuronal benchmarks, applying CellProfiler, Cytoself, and Paired Cell Inpainting to CellAligner-mapped images rather than to the original images improved the ability of these methods to distinguish ground-truth localization patterns by more than two-fold.

100

Pablo G. Camara @camarapg.bsky.social · 12/06/2026

These latent spaces can then be analyzed with standard machine learning tools for clustering, dimensionality reduction, multimodal integration, trajectory inference, enrichment analysis, etc.

100

Pablo G. Camara @camarapg.bsky.social · 12/06/2026

CellAligner produces both morphological and subcellular localization latent spaces. In these spaces, points represent cells, and pairwise distances represent morphological and subcellular organization distances, respectively.

100

Pablo G. Camara @camarapg.bsky.social · 12/06/2026

Mapping is guided by morphological stains, so that the resulting maps preserve compartmental structure rather than simply warping images in an unconstrained way. Being unsupervised, it does not require training data and can be applied to small datasets.

100

Pablo G. Camara @camarapg.bsky.social · 12/06/2026

CellAligner uses fused unbalanced Gromov-Wasserstein couplings to map protein imaging data across morphologically diverse cells. This allows subcellular localization patterns to be compared within a common reference cellular morphology.

100

Pablo G. Camara @camarapg.bsky.social · 12/06/2026

We introduce CellAligner, an unsupervised framework for studying subcellular organization across cells with complex, heterogeneous shapes. In these settings, image-analysis methods can be strongly affected by cell shape, making it hard to identify true localization differences.

100

Reposted by Pablo G. Camara

TanentzapfLab @tanentzapflab.bsky.social · 13/08/2025

The goal of a PhD is not to learn some facts or read a few papers or learn a bunch of techniques. The goal of a PhD is to learn independence, problem solving, how to finish things you start, resilience, & gain the ability to adapt & think creatively. Learning these things is hard.

433587

Reposted by Pablo G. Camara

Michael P Hart 🏳️‍🌈 (he/him) @michaelphart1.bsky.social · 13/08/2025

Do you like neuronal #plasticity? #genetics? #Celegans? Check out 1st installment of work by Kristi Zoga in the lab. Kristi screened 20 conserved #autism genes for roles in experience-dependent #neuron remodeling and a related behavior. doi.org/10.1093/genetics/iyaf156

doi.org

Multiple autism genes influence GABA neuron remodeling via distinct developmental trajectories

Abstract. Variation in over 100 genes are now associated with increased risk for autism and related neurodevelopmental conditions, but how this variation r

3307