cosmx

Production-grade multimodal deep learning for biological imaging and spatial transcriptomics, with reproducible data ingestion, image quality control, segmentation, phenotype classification, uncertainty estimation, and transcriptomic interpretation.

Python toolkit for image-based spatial transcriptomics (CosMx, MERFISH, Xenium). Unified data object for expression, polygons, and FOV images, plus QC, clustering, point- and polygon-based spatial statistics, and polygon-native cell-cell communication using real membrane contact geometry.

Transcript-informed cell segmentation for imaging spatial transcriptomics (Xenium, CosMx, MERSCOPE). Bayesian cell boundary sampling after Jones et al., Nat Methods 2025. A Go library and a single self-contained command.

Signed multi-module composite activity scoring for CosMx SMI spatial transcriptomics, with per-cell scores registered onto tissue images

Workflows and code for spatial and meta-transcriptomic analysis of seronegative hepatitis, autoimmune hepatitis and donor liver samples.

End-to-end R/Seurat pipeline for CosMx SMI spatial transcriptomics: memory-safe object handling, SCTransform, Louvain clustering, InSituType cell typing, and spatialTIME nearest-neighbour G statistics.

single-cell RNA-seq, spatial transcriptomics, and spatial proteomics analysis skills

Analyze biological imaging and spatial transcriptomics data using robust deep learning pipelines and verified dataset auditing.