Drug-induced mRNA Signatures Are Enriched for the Minority of Genes that Are Highly Heritable

Tianxiang Gao, Petter Brodin, Mark M. Davis, Vladimir JojicPacific Symposium on Biocomputing 2015

Abstract

A central challenge in drug discovery is determining whether therapeutic targets should be genes influenced primarily by genetics (nature) or the environment (nurture). This study integrates twin gene expression data, drug-response signatures, and regulatory network modeling to address this question. We find that while most genes in human blood are driven by environmental factors, drug-induced expression changes are significantly enriched for the small minority of highly heritable genes. Furthermore, we demonstrate that "heritable flow strength" within gene regulatory networks is a powerful predictor of drug influence, whereas environmental noise is a poor predictor.

Introduction

When identifying drug targets, researchers often utilize blood gene expression signatures as biomarkers for disease. However, it remains unclear if drugs should target genes that are "flexible" due to environmental sensitivity or those "locked-in" by genetic background. This research investigates the relationship between a gene's heritability and its propensity to be influenced by pharmacological agents, challenging the intuition that environmental flexibility makes a gene a better target.

Methodology

The researchers developed a multi-stage modeling pipeline to map the interplay between genetics, networks, and drug responses.

Integrated modeling workflow combining drug-gene relations, twin expression data, and regulatory networks.

The approach utilized three primary components:

  1. Twin-ACE Modeling: Using the MuTHER dataset (blood samples from 718 twins), the authors decomposed gene expression variance into Additive genetic effects (A), Common environmental effects (C), and unique Environmental effects (E).
  2. Network Construction: A Directed Acyclic Gene Regulatory Network (DAGRN) was built using 257 human signaling pathways from the KEGG database.
  3. Flow Decomposition: The team calculated "regulation flow" between genes, further decomposing it into Heritable Flow Strength (propagation of genetic variance) and Non-heritable Flow Strength (propagation of environmental noise).

Results and Key Insights

The study revealed a fundamental "Heritability Paradox." As shown in the distribution of heritability across the genome, the vast majority of genes are dominated by non-heritable factors.

Distribution of gene heritability across the MuTHER dataset, showing that most genes are primarily influenced by environmental factors.

Despite this environmental dominance, drugs disproportionately affect the small subset of genes with high heritability. Key findings include:

  • Target Preference: Using a strict significance threshold (p < 0.001), the researchers identified four drugs (e.g., Trichostatin A and Dobutamine) that target heritable genes, while only one drug (Mefenamic acid) showed a preference for non-heritable targets.
  • Predictive Power of Heritable Flow: When predicting whether a drug affecting a "source" gene would also influence a "target" gene, Heritable Flow Strength achieved an Area Under the Curve (AUC) of 0.63. In contrast, total Regulation Flow Strength achieved an AUC of 0.44, and Non-heritable Flow Strength performed worse than random guessing (AUC = 0.38).

These results suggest that drugs do not "hijack" the pathways used by environmental fluctuations; instead, they follow the structured pathways established by genetics.

Discussion and Future Directions

While these findings provide a new compass for drug discovery, several limitations remain. The use of Directed Acyclic Graphs (DAGs) simplifies biological systems by excluding feedback loops. Additionally, because the study relied on blood samples, the findings may vary in other tissues like the liver or brain. Future work should focus on expanding these models to include complex network architectures and tissue-specific data to refine the prediction of drug influence flows and minimize off-target effects.

Summary updated

BibTeX
@inproceedings{gao2015drug,
  author = {Tianxiang Gao and Petter Brodin and Mark M. Davis and Vladimir Jojic},
  title = {{Drug-induced mRNA Signatures Are Enriched for the Minority of Genes that Are Highly Heritable}},
  booktitle = {Pacific Symposium on Biocomputing},
  year = {2015},
}
Figure