MANGO (MAppiNG multi-cell-type multi-scale spatial relationships in spatially resolved Omics)
MANGO is a permutation-based computational framework for detecting higher-order spatial co-localization across multiple spatial scales.
MANGO defines the local microenvironment of each reference cell using a k-nearest-neighbor neighborhood with Gaussian distance-decay weighting. Within each neighborhood, MANGO computes a multivariate local co-localization quotient (mLCLQ) that quantifies whether a predefined set of query cell types co-occurs more frequently than expected given their global spatial distributions. Significance is assessed through permutation testing. MANGO is implemented as an open-source R package, available at: https://github.com/ZihanYang030/mango.
Schematic of mLCLQ calculation
For each reference cell, we compute local weighted proportions of the query types within neighborhoods using spatial weights and observed cell-type labels. The proportions are combined and divided by the global cell-type abundance to obtain an observed mLCLQ score. We then repeat the calculation with independently permuted labels produces null scores.
MANGO is under active development. We appreciate any suggestions and feedback. Contact: zy56@rice.edu.
Install MANGO from GitHub:
remotes::install_github("ZihanYang030/MANGO")
library(MANGO)MANGO is licensed under the GNU General Public License v3.0 or any later version.