Novel methods for detecting positive selection in mosquitoes

/Abstract Rapid adaptation has been implicated in a wide range of biological processes relating to human health, ranging from antibiotic resistance in bacteria to insecticide resistance in mosquitoes. However, accurately detecting the genomic signatures of rapid adaptation remains a challenging task in evolutionary biology that has been historically limited by a lack of genomic data and sufficiently accurate methods for analyzing such data. Machine learning tools offer great promise for population genetics due to their unrivaled capacity for detecting high-dimensional signatures of natural selection and their ability to account for demographic histories which may produce signatures that mimic that of selection. Nonetheless, even modern machine learning tools suffer from two significant limitations: 1) their capacity to distinguish between hard and soft sweeps which is vital for understanding the pace that an organism can adapt and 2) they can experience drops in accuracy in the presence of severe model mis-specification (e.g., when the true demographic history of a population is unknown or incorrectly inferred). Indeed, there is no method capable of accounting for both limitations simultaneously. The proposed research will first develop a method capable of detecting and distinguishing between multiple modes of selection while explicitly accounting for model mis-specification, and then apply this method to populations of the yellow fever mosquito to explicate the genomic underpinnings of rapid adaptation to insecticides (Aim 1). Next, long-read sequencing data will be used to explore the role of structural variation (genomic variation >50bp) in rapid adaptation (Aim 2). The final aim will develop the first machine learning tool to detect an understudied form of rapid adaptation, namely adaptive tracking, where fluctuating selective pressures result in repeatable adaptive shifts in allele frequencies over seasonal and sub-seasonal timescales. A multi-year field-based experiment will then be conducted where the southern house mosquito will be exposed to both oscillating environmental and insecticide-based directional selection. The novel machine learning tool will then be applied to this dataset to characterize the targets of fluctuating selection in a major disease vector and assess whether the presence of strong direction selection alters the degree, tempo, and targets of adaptive tracking (Aim 3). Dr. Ketchum has assembled a team of expert mentors who will help broaden her knowledge in machine learning, insect genomics, and mosquito rearing protocols. Dr. Ketchum’s primary mentor, Dr. Dan Schrider has pioneered some of the first applications of machine learning tools to population genetic datasets and so is perfectly suited to help Dr. Ketchum achieve her research goals. The K99 phase of the award will take place within the Department of Genetics at UNC Chapel Hill which is an intellectually stimulating environment with ample opportunities to participate in journal clubs and seminar series and collaborate with other research groups. This training will help Dr. Ketchum successfully complete her proposed research and aid her transition to principal investigator of an internationally recognized lab that studies the genomic architecture of adaptation. Project Number: 1K99GM160783-01A1 | Fiscal Year: 2026 | NIH Institute/Center: National Institute of General Medical Sciences (NIGMS) | Principal Investigator: Remi Ketchum | Institution: UNIV OF NORTH CAROLINA CHAPEL HILL, CHAPEL HILL, NC | Award Amount: $113,464 | Activity Code: K99 | Study Section: Special Emphasis Panel[ZRG1 MGG-M (80)] View on NIH RePORTER: https://reporter.nih.gov/project-details/11372389