Abstract
The discovery of genetic risk factors has transformed human genetics, yet the pace of new gene identification has slowed despite the exponential expansion of sequencing and biobank resources. Current approaches are optimized for the extremes of the allele frequency spectrum: rare, high-penetrance variants identified through burden testing and common, low-effect variants mapped by genome-wide association studies. Between these extremes lie variants of intermediate frequency and effect size for which statistical power is limited, pathogenicity is often misclassified, and gene discovery lags behind empirical evidence of heritable contribution. This "missing middle" represents a critical blind spot across disease areas, from neurodevelopmental and psychiatric disorders to cancer and aging. In this review, we organize strategies for risk gene identification by variant frequency class, highlighting methodological strengths and constraints at each scale. We draw on lessons across fields to illustrate how innovations in variant annotation, joint modeling, phenotype refinement, and network-based inference can extend discovery into the intermediate range. By framing the frequency spectrum as a unifying axis, we provide a conceptual map of current capabilities, their limitations, and emerging directions toward more comprehensive risk gene discovery.