# SM3DD with segmented PCA: a comprehensive method for interpreting 3D spatial transcriptomics

## TL;DR

The paper introduces SM3DD, a new method for analyzing 3D spatial RNA data to compare lung tissues from healthy individuals and SARS-CoV-2 patients.

## Contribution

SM3DD is a novel, segmentation-free approach that enables pathway discovery in spatial transcriptomics without requiring cell annotation.

## Key findings

- SM3DD identified significant gene distance changes in SARS-CoV-2 patients compared to normal individuals.
- Hierarchical clustering of SM3DD results grouped genes by function, aiding biological interpretation.
- Segmented PCA revealed pathways like 'SARS-CoV-2 infection' despite no viral transcripts being measured.

## Abstract

We developed Standardised Minimum 3D Distance (SM3DD), an entirely cell segmentation/annotation-free approach to the analysis of spatial RNA datasets, using it to compare lung tissue from 16 clinically normal individuals to that of 18 SARS-CoV-2 patients who died from acute respiratory distress syndrome. RNA spatial coordinates were determined using the CosMx™ Spatial Molecular Imager (Bruker Spatial Biology, US). For each individual transcript location, we calculated the three-dimensional distances to the nearest transcript of each transcript type, standardising the distances to each transcript type. Mean SM3DDs were compared between normal and SARS-CoV-2 patients. Notably, hierarchical clustering of the directional log10(P) values organized genes by functionality, making it easier to interpret biological contexts, and for FKBP11, where a decrease in distance to MZT2A was the most significant difference, suggesting a role in interferon signalling. Using a segmented principal components analysis of the entire SM3DD dataset, we identified multiple pathways, including ‘SARS-CoV-2 infection’, even though the assay did not include any SARS-CoV-2 transcripts.