A Network-Structured Bayesian Hierarchical Model for Sparse Mutation-Drug Response Associations: Application to Cancer Pharmacogenomics
Abstract
We develop a network-structured Bayesian hierarchical model for sparse association mapping between genomic alterations and quantitative treatment-response phenotypes. The framework combines a Gaussian Markov random field prior that borrows strength across pathway-connected genes, a global-local horseshoe prior inducing sparsity, and a conjugate Gibbs sampler requiring no Metropolis-Hastings steps. Though broadly applicable to high-dimensional settings with known predictor networks, we validate it using cancer cell-line drug-sensitivity data. Applied to GDSC2 (N=951 cell lines, G=219 driver genes, D=295 drugs), the model identifies 126 gene-drug associations (0.195\% of 64{,}605 pairs), concentrated in EZH2 (45 drugs, all sensitivity-direction, mean effect −0.911 lnIC50) and KMT2D (36 drugs, all sensitivity-direction, mean effect −0.496 lnIC50). These markers show external support in an independent PRISM screen (1{,}518 compounds), with KMT2D achieving complete directional replication (36/36) and EZH2 partial replication (8/12). Five-fold cross-validated predictive log-likelihood confirms each prior layer's value: the full model outperforms the no-network ablation by +3,109 log-units per fold and the no-horseshoe ablation by +14,039 log-units, consistently across folds. Simulations under three scenarios show the full model achieves the highest precision and lowest false-discovery rate throughout, while the network prior improves sensitivity recovery under network-structured signal. A tissue-stratified extension identifies coherent subgroup refinements, including lung-specific EGFR-inhibitor sensitivity and skin-specific BRAF-Dabrafenib sensitivity. These results show the framework identifies sparse, interpretable, externally supported drug-sensitivity markers while enabling principled investigation of tissue-specific departures from shared effects.
Related
- Comparative evolutionary pharmacogenomics of human prostaglandin-endoperoxide synthase paralogs identifies population-structured coding variation and protein-contextual candidates in N-terminal leader-sequence and catalytic-channel regions
- A hierarchical Bayesian model for dynamic phenotypes in multi-environment trials with genotype by environment interaction
- Mitigating Class Imbalance in Graph-Structured Data via Hierarchical Learning: Insights from Protein Binding Site Prediction
- Bayesian Hierarchical Models for Longitudinal Meta-Analysis in Sports Science: A Practical Tutorial Using brms
- Behavioral Mapping of Hierarchical Dynamical Models Accelerates Bayesian Parameter Inference
BibTeX
@misc{olayinka2026networkstructured,
title = {{A Network-Structured Bayesian Hierarchical Model for Sparse Mutation-Drug Response Associations: Application to Cancer Pharmacogenomics}},
author = {Hammed A. Olayinka and Saheed O. Olayemi},
howpublished = {arXiv},
year = {2026},
month = sep,
doi = {10.48550/arxiv.2609.05784},
eprint = {2609.05784},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2609.05784},
}
RIS
TY - UNPB
TI - A Network-Structured Bayesian Hierarchical Model for Sparse Mutation-Drug Response Associations: Application to Cancer Pharmacogenomics
AU - Hammed A. Olayinka
AU - Saheed O. Olayemi
PY - 2026
DA - 2026/09/05
PB - arXiv
DO - 10.48550/arxiv.2609.05784
UR - https://arxiv.org/abs/2609.05784
AB - We develop a network-structured Bayesian hierarchical model for sparse association mapping between genomic alterations and quantitative treatment-response phenotypes. The framework combines a Gaussian Markov random field prior that borrows strength across pathway-connected genes, a global-local horseshoe prior inducing sparsity, and a conjugate Gibbs sampler requiring no Metropolis-Hastings steps. Though broadly applicable to high-dimensional settings with known predictor networks, we validate it using cancer cell-line drug-sensitivity data. Applied to GDSC2 (N=951 cell lines, G=219 driver genes, D=295 drugs), the model identifies 126 gene-drug associations (0.195\% of 64{,}605 pairs), concentrated in EZH2 (45 drugs, all sensitivity-direction, mean effect −0.911 lnIC50) and KMT2D (36 drugs, all sensitivity-direction, mean effect −0.496 lnIC50). These markers show external support in an independent PRISM screen (1{,}518 compounds), with KMT2D achieving complete directional replication (36/36) and EZH2 partial replication (8/12). Five-fold cross-validated predictive log-likelihood confirms each prior layer's value: the full model outperforms the no-network ablation by +3,109 log-units per fold and the no-horseshoe ablation by +14,039 log-units, consistently across folds. Simulations under three scenarios show the full model achieves the highest precision and lowest false-discovery rate throughout, while the network prior improves sensitivity recovery under network-structured signal. A tissue-stratified extension identifies coherent subgroup refinements, including lung-specific EGFR-inhibitor sensitivity and skin-specific BRAF-Dabrafenib sensitivity. These results show the framework identifies sparse, interpretable, externally supported drug-sensitivity markers while enabling principled investigation of tissue-specific departures from shared effects.
ER -