Determining cellular phylogeny and migration at high resolution in spatial genomics

Cellular migration and tumor evolution play crucial roles in tissue homeostasis and cancer biology. Spatial genomics technologies have opened up the potential to track these processes in situ within human tissues using somatic mutations. However, the inherent data dropouts in spatial genomics measurements pose a challenge to this approach.

In this study, the authors introduce StarTrak, a framework designed to address these challenges by enabling somatic variant calling and lineage reconstruction at single-pixel spatial resolution.

StarTrak utilizes a neural network framework to integrate multiple types of data, including mutational evidence from both nuclear and mitochondrial reads, spatial proximity, histological images, and transcriptomic states. This comprehensive approach allows StarTrak to identify spatially informative patterns of mutations. The framework has shown robust performance across various tissue types and spatial-omics platforms, with particularly strong results when applied to spatial ATAC-seq data.

When applied to tumor sections, StarTrak was able to infer the migration of mutated tumor cells into adjacent non-malignant tissues. The framework also successfully reconstructed distinct lineage trees associated with different spatial regions. An interesting finding was the identification of tumor regions that exhibited nearly identical transcriptomic states but had distinct mutational profiles. This highlights the potential of StarTrak to uncover the complex interplay between mutations and cellular states within tumors.

Furthermore, StarTrak demonstrated its ability to identify sunlight-associated mutations in non-malignant epidermal cells within skin sections. This suggests that the framework may also be useful in understanding the effects of environmental factors on cellular mutations. Additionally, the study found evidence suggesting that these mutant surface cells might migrate into deeper skin layers, a finding that could have implications for understanding skin cancer development and progression.

In conclusion, StarTrak represents a significant advancement in the field of spatial genomics. By integrating mutation calling with spatial context, the framework enables the inference of spatially resolved cellular migration and phylogeny. This provides valuable insights into the complex processes underlying tissue homeostasis, cancer evolution, and the impact of environmental factors on cellular mutations.