Abstract

   Spatial transcriptomics (ST) technology through in situ capturing has
   enabled topographical gene expression profiling of tumor tissues.
   However, each capturing spot may contain diverse immune and malignant
   cells, with different cell densities across tissue regions. Cell type
   deconvolution in tumor ST data remains challenging for existing methods
   designed to decompose general ST or bulk tumor data. We develop the
   Spatial Cellular Estimator for Tumors (SpaCET) to infer cell identities
   from tumor ST data. SpaCET first estimates cancer cell abundance by
   integrating a gene pattern dictionary of copy number alterations and
   expression changes in common malignancies. A constrained regression
   model then calibrates local cell densities and determines immune and
   stromal cell lineage fractions. SpaCET provides higher accuracy than
   existing methods based on simulation and real ST data with matched
   double-blind histopathology annotations as ground truth. Further,
   coupling cell fractions with ligand-receptor coexpression analysis,
   SpaCET reveals how intercellular interactions at the tumor-immune
   interface promote cancer progression.

   Subject terms: Cancer microenvironment, Computational models, Cancer
   genomics, Tumour heterogeneity, Software
     __________________________________________________________________

   Cell type deconvolution in tumor spatial transcriptomics (ST) data
   remains challenging. Here, the authors develop Spatial Cellular
   Estimator for Tumors (SpaCET) to infer cell types and intercellular
   interactions from ST data in cancer across different platforms, with
   improved performance over similar methods.

Introduction

   Profiling the transcriptome of cells in their spatial context is
   critical to a mechanistic understanding of tumor progression and
   therapeutic resistance^[32]1. Recent years have seen the rapid
   development of spatial transcriptomics (ST) with gene coverage from a
   few targets to genome-wide and various cellular resolutions from
   subcellular to multiple cells^[33]2,[34]3. As a key branch of ST
   methods, in situ capturing strategy based on positional molecular
   barcodes enables unbiased capture of the whole transcriptome within
   intact tissue^[35]3. Its representative techniques include
   Slide-seq^[36]4, 10x Visium^[37]5, and the early in situ capturing
   method from which Visium was developed^[38]6. Specifically, the
   commercial Visium platform can profile mRNA levels in fresh-frozen and
   formalin-fixed paraffin-embedded (FFPE) tissues, enabling their
   widespread application^[39]7. However, the spatial spot of various
   capturing strategies with a 10–100 μm diameter might measure a mixture
   of signals from multiple cells of different lineages. Consequently,
   decomposing cell identities in spots is a critical step in
   characterizing the spatial cellular landscape of tissues.

   Many methods exist for cell type decomposition in general ST
   data^[40]8–[41]13 and bulk transcriptome profiling^[42]14–[43]17.
   However, it is challenging for these methods and their underlying
   strategies to address the unique issue of tumor ST data. Several
   methods, such as Stereoscope^[44]8, RCTD^[45]11, and CIBERSORTx^[46]15,
   predict cancer cell fractions relying on the availability of suitable
   malignant reference profiles. Other methods, such as EPIC^[47]14,
   estimate malignant cell fraction without references by estimating the