Selected work

Representative Papers

Combining Bayesian and Frequentist Inference for Laboratory-Specific Performance Guarantees in Copy Number Variation Detection

Apr 15, 2026

This study addresses the challenges in clinically validating copy number variation (CNV) detection from targeted amplicon sequencing, where performance is hindered by amplification artifacts, heterogeneity from protocol mismatches, and limited sample sizes. The authors propose a hybrid framework integrating Bayesian and frequentist inference: Bayesian posterior functionals are employed to assess performance, with squared loss modeled via a Gamma distribution to construct admissible intervals achieving valid frequentist coverage. Innovatively, the method incorporates a label-free mechanism to exclude CNV-positive outliers, small-sample regularization, and a log-model-evidence–based stratification strategy to effectively mitigate non-exchangeable noise. Evaluated on two amplicon panels, the approach achieves single-digit mean absolute coverage error across all genes under both protocol-matched and mismatched conditions, substantially outperforming conventional Bayesian methods—for instance, reducing ERBB2 error by over 60%.

0 citationsRead paper

Detecting Batch Heterogeneity via Likelihood Clustering

Jan 14, 2026

This study addresses the challenge in genomic diagnostics where batch effects are often confounded with biological signals such as copy number variations (CNVs), leading to false positives or missed detections—particularly when batch labels are unavailable. The authors propose a novel Bayesian approach that requires no prior batch information and instead leverages model evidence to cluster samples: technical artifacts reduce model evidence, whereas genuine biological variation does not. Heterogeneity is identified via a likelihood ratio test in evidence space, calibrated using a parametric bootstrap procedure. This work represents the first application of model evidence to disentangle technical from biological signals. The method demonstrates superior clustering accuracy over correlation- and dimensionality-reduction-based approaches on synthetic data, three clinical targeted sequencing panels (liquid biopsy, BRCA, and thalassemia), and mouse electrophysiology data, while maintaining strict control over false positive rates.

0 citationsRead paper

Recent publications

Latest Papers

Combining Bayesian and Frequentist Inference for Laboratory-Specific Performance Guarantees in Copy Number Variation Detection

Apr 15, 2026

This study addresses the challenges in clinically validating copy number variation (CNV) detection from targeted amplicon sequencing, where performance is hindered by amplification artifacts, heterogeneity from protocol mismatches, and limited sample sizes. The authors propose a hybrid framework integrating Bayesian and frequentist inference: Bayesian posterior functionals are employed to assess performance, with squared loss modeled via a Gamma distribution to construct admissible intervals achieving valid frequentist coverage. Innovatively, the method incorporates a label-free mechanism to exclude CNV-positive outliers, small-sample regularization, and a log-model-evidence–based stratification strategy to effectively mitigate non-exchangeable noise. Evaluated on two amplicon panels, the approach achieves single-digit mean absolute coverage error across all genes under both protocol-matched and mismatched conditions, substantially outperforming conventional Bayesian methods—for instance, reducing ERBB2 error by over 60%.

0 citationsRead paper

Detecting Batch Heterogeneity via Likelihood Clustering

Jan 14, 2026

This study addresses the challenge in genomic diagnostics where batch effects are often confounded with biological signals such as copy number variations (CNVs), leading to false positives or missed detections—particularly when batch labels are unavailable. The authors propose a novel Bayesian approach that requires no prior batch information and instead leverages model evidence to cluster samples: technical artifacts reduce model evidence, whereas genuine biological variation does not. Heterogeneity is identified via a likelihood ratio test in evidence space, calibrated using a parametric bootstrap procedure. This work represents the first application of model evidence to disentangle technical from biological signals. The method demonstrates superior clustering accuracy over correlation- and dimensionality-reduction-based approaches on synthetic data, three clinical targeted sequencing panels (liquid biopsy, BRCA, and thalassemia), and mouse electrophysiology data, while maintaining strict control over false positive rates.

0 citationsRead paper