Applied physics approaches for the construction of an integrated nanostring GeoMx atlas of the healthy mouse brain

Spatial transcriptomics enable the simultaneous measurement of gene expression and spatial context, producing high-dimensional datasets able to capture tissue architecture and cellular heterogeneity. Physics provides the conceptual and mathematical tools required to describe such complex systems. In this work we developed an integrated reference atlas of the mouse brain using Nanostring GeoMx DSP data by combining three independent datasets. The pipeline comprises robust quality control, normalization procedure, dataset integration and batch correction. These operations can be interpreted as extracting the ‘ground state’ of a system, free from any noise or external perturbations. Clustering and variance metrics provide quantitative measures of system order and biological concordance. DESeq2 emerged as the best normalization method thanks to its ability to model library size and composition bias through size-factor estimation. Regarding batch correction, Limma was selected as the best performing method, managing to account even for within-dataset technical variability and unbalanced batches. Dimensionality reduction techniques, such as PCA and UMAP, further highlight the emergence of coherent patterns from high-dimensional noise, similar to uncovering latent modes in statistical mechanics. Lastly, the integrated atlas preserves region-specific gene expression while increasing the statistical power to detect differential expression, demonstrating the validity of the approach. Overall, we can consider this framework generalizable to other high-dimensional datasets, offering a transferable strategy for integrating, denoising and interpreting complex biological systems. Future extensions may incorporate network-theoretical or energy-based models to map dynamic interactions and energy landscapes within tissues, bridging quantitative biology and applied physics, while enabling improved stratification of the atlas on additional biological variables such as age or sex.