Researchers at Marquette University have developed MIRAGE, a new Bayesian computational method that addresses the challenges of rare genetic variant association tests. Standing for “mixture-model-based rare-variant analysis on genes,” MIRAGE was found to better account for the heterogeneity of variant effects by focusing on each single variant through Bayesian approach.
Rare genetic variants play a substantial role in conferring disease risk, but not all rare variants are risk variants and identifying which are true risk variants (usually very few ) is challenging. MIRAGE was developed to address this issue of detecting the sparse number of rare risk variants associated with diseases by leveraging their functional annotations.
Dr. Shengtong Han, assistant professor of biostatistics in Marquette’s School of Dentistry, published the findings in the January 2026 issue of The American Journal of Human Genetics, “MIRAGE: A Bayesian statistical method for gene-level rare-variant analysis incorporating functional annotations.” Dr. Wenhui Sheng, former research assistant professor of mathematical and statistical sciences at Marquette, was among the paper’s co-authors.
As a demonstration, Han and his co-authors tested MIRAGE using exome data, i.e. protein-coding parts of DNA, of autism.
MIRAGE analyzes summary statistics, such as variant counts from inherited variants in trio sequencing or from ancestry-matched case-control samples. It then captures the heterogeneity of variant effects by treating all variants of a gene as a mixture of risk and non-risk variants and uses external information of variants to model the prior probabilities of being risk variants. The authors demonstrated, in both simulations and analysis of an exome-sequencing dataset of autism, that MIRAGE significantly outperforms current methods for rare-variant analysis.
“Rare genetic variant association test is challenging due to sparse number of variants and the heterogeneity of variant effects,” Han said. “We proposed a Bayesian method, MIRAGE, to address these challenges. Our findings on autism studies by MIRAGE indicate that risk genes could be contributed by variants with varying effect sizes. Furthermore, MIRAGE is able to detect risk genes with very few numbers of large effect variants.”
Han and his colleagues successfully applied MIRAGE to autism whole exome sequence (WES) data, providing new insights into the risk genes associated with autism. The top genes MIRAGE identified were highly enriched with known or plausible autism-risk genes. When applying MIRAGE to autism WES data, risk genes were prioritized. The signals of top risk genes are from different variant groups, highlighting the heterogeneous variant effects.
Dr. Xin He of the University of Chicago was senior author on the paper, providing guidance and funding support. Additional author contributions include Dr. Xiaotong Sun, a postdoctoral researcher in the He lab; Laura Sloofman, bioinformatician at Icahn School of Medicine at Mount Sinai; and Dr. Kyle Satterstrom, a computational biologist at the Broad Institute of MIT and Harvard.
