CELESTA addresses a core need in spatial biology and cancer research: assigning cell types in multiplexed in situ imaging data without requiring labeled training data. This open-source R package jointly models each cell’s protein expression profile and its spatial neighborhood context via an iterative EM algorithm with Markov Random Field priors, supporting scalable and reproducible cell type mapping directly in tissue sections. It’s a natural fit for studies characterizing tumor immune microenvironments and discovering spatially organized cell communities linked to clinical outcomes such as metastasis.
Sustainability Snapshot via Cancer Complexity Toolkit:
📚 Comprehensive documentation with README, dependency tracking, and worked example data included in the repository
🔧 Active development on GitHub (plevritis-lab/CELESTA)
✅ Quality indicators: JOSS Score 0.32 | Almanack 0.45
🌱 All four Almanack checks pass — one of the few CCKP tools
Why this matters: CELESTA earns a “Foundational” Almanack grade (0.45) with a notable distinction: it’s one of the only tools in the CCKP registry where all four sustainability checks pass (with complete README, dependency presence, test coverage). For labs integrating multiplexed imaging data into cancer research workflows, it provides a reproducible, well-tested path from raw CODEX or similar imaging data to spatially resolved cell type maps and tissue architecture insights.
Attribution: Developed by the Plevritis Lab, Department of Radiology, School of Medicine, Stanford University (plevritis-lab/CELESTA).
