Bayesian Copula Directional Dependence is Cross-Network Robust for Gene-Regulatory Pair Direction: A Benchmark Study on DREAM5

2026-06-28

AI Summary

Existing methods struggle to reliably infer gene regulatory directions under cross-network settings (from synthetic to real organisms) and in few-shot scenarios. This work proposes a Bayesian Copula Directional Dependence (CDD) framework that enables uncertainty-aware directional refinement through posterior distributions, 95% credible intervals, and sign-support scores, while incorporating a non-decision mechanism to abstain from low-confidence predictions. Evaluated on the three DREAM5 networks—spanning both prokaryotic and eukaryotic organisms—the method achieves, for the first time, consistently above 60% accuracy and 88% coverage without performance degradation in small-sample regimes. Comprehensive benchmarking demonstrates that it significantly outperforms eight existing approaches in accuracy, coverage, and directional AUROC, standing as the only Bayesian method that simultaneously maintains high coverage and substantially exceeds random guessing across all evaluated scenarios.

Abstract

Inferring the direction of a gene-regulatory relationship is harder than inferring whether a relationship exists, and most direction-inference methods are validated mainly on a single in silico benchmark. We ask which method remains reliable as the data move from a synthetic network to real organisms and as sample size decreases. We embed a copula-based measure of directional dependence (CDD) in a Bayesian framework that returns, for each candidate pair, a posterior distribution over a directional contrast, a 95% credible interval, a posterior sign-support score, and a principled no-call. We benchmark this estimator against eight direction-inference methods, including two Bayesian DAG-posterior baselines, on the three core DREAM5 networks (in silico, S. aureus, and E. coli), with S. cerevisiae used as an out-of-domain eukaryotic stress test. Across the three core networks, Bayesian CDD is the only method whose called accuracy is always above 60%, whose coverage is always above 88%, and whose direction AUROC is always above 0.6; every competing method falls to chance or below on at least one network. CDD ranks first on both real-organism networks, remains stable on the smallest-sample network where bootstrap-interval methods collapse, and is the only Bayesian method that is simultaneously above chance and high-coverage under a 95% posterior gate. We position CDD as a post-screening, uncertainty-aware direction-refinement tool for candidate regulatory pairs.

Problem

• gene-regulatory direction
• directional dependence
• cross-network robustness
• Bayesian inference
• DREAM5 benchmark

Innovation

• Bayesian Copula Directional Dependence
• gene regulatory network
• direction inference
• cross-network robustness
• uncertainty quantification