SpaPheno: Linking Spatial Transcriptomics to Clinical Phenotypes with Interpretable Machine Learning

Overview

Linking spatial transcriptomic patterns to clinically relevant phenotypes is a critical step toward spatially informed precision oncology. Here, we introduce SpaPheno, an interpretable machine learning framework that integrates spatial transcriptomics with clinically annotated bulk RNA-seq data to uncover spatially resolved biomarkers predictive of patient outcomes. Leveraging Elastic Net regression combined with SHAP-based attribution, SpaPheno uniquely identifies spatial features at multiple scales—from tissue regions to specific cell types and individual spatial spots—that are associated with patient survival, tumor stage, and immunotherapy response. We demonstrate the robustness and generalizability of SpaPheno through comprehensive simulations and applications spanning primary liver cancer, clear cell renal cell carcinoma, breast cancer, and melanoma. Across these diverse settings, SpaPheno achieves high predictive accuracy while providing biologically meaningful and spatially precise interpretations. Our framework offers a powerful and extensible approach for translating complex spatial omics data into actionable clinical insights, accelerating the development of precision oncology strategies grounded in tumor spatial architecture.

The Overview of SpaPheno

☀️ Key Features

• Integration of spatial transcriptomics with clinically annotated bulk RNA-seq data

• Multi-scale interpretable machine learning framework

• Robust applicability across diverse cancer types and clinical endpoints

⏬ Installation

if (!require("BiocManager", quietly = TRUE)) {
  install.packages("BiocManager")
}

## Install suggested packages
# BiocManager::install(c(
#   "glmnet",
#   "FNN",
#   "survival"
# ))

# install.packages("devtools")
# devtools::install_github("bm2-lab/SpaDo")

# SpaPheno installation
# devtools::install_github("Duan-Lab1/SpaPheno", dependencies = c("Depends", "Imports", "LinkingTo"))

library(SpaPheno)
library(tidyverse)
library(ggplot2)
library(reshape2)
library(stringr)
library(survival)

Download the pre-packaged installation package directly from the GitHub repository

install.packages("SpaPheno_0.0.1.tar.gz", repos = NULL, type = "source")

🚀 Quick Start

Data availability

The data required for the test are all listed in the following google cloud directory SpaPheno Demo Data.

├── BRCAsurvival.RData
├── HCC_stage.RData
├── HCC_survival.RData
├── KIRC_survival.RData
├── Melanoma_ICB.RData
├── Simulation_osmFISH.RData
└── Simulation_STARmap.RData

Tutorial

For more information and documentation, please visit the SpaPheno website.

📖 Vignette

Using the following command and Choosing the html for more details.

utils::browseVignettes(package = "SpaPheno")

💖 Contributing

Welcome any contributions or comments, and you can file them here.

🏆 Acknowledgement

Thanks all the developers of the methods integrated into SpaPheno.

✴️ Citation

Kindly cite by using citation("SpaPheno") if you think SpaPheno helps you. Alternative way is Bin Duan (2025). SpaPheno: Linking Spatial Transcriptomics to Clinical Phenotypes with Interpretable Machine Learning. R package version 0.0.1, <URL:https://github.com/Duan-Lab1/SpaPheno>.

✍️ Authors

• Bin Duan

• Hua Zou