Research

EVOLUTIONARY INFERENCE USING MACHINE LEARNING

Illustration of a genome alignment used as input to a neural network to infer the locations of selective sweeps.

Machine learning (ML) for evolutionary inference has been shown to outperform traditional approaches. My research aims to further develop ML tools to study processes such as adaptation and to bring these tools to non-model organisms. 

Traditional approaches for detecting recent positive selection often focus on classic “hard” selective sweeps, yet adaptation can also occur through standing genetic variation or recurrent adaptive mutations, producing “soft” sweeps. These approaches also struggle to discriminate between the effects of demographic events from selective sweeps. I leverage the superior accuracy of ML-based approaches to distinguish between these different modes of selection and elucidate novel regions of the genome under selection in disease-vector mosquitoes. 

In my most recent work I found that soft selective sweeps are substantially more common than hard sweeps in the yellow fever mosquito, suggesting that these mosquitoes are able to adapt to stressors rapidly as they need not 'wait' for a de novo mutation to arise. I identified both well-characterized and putatively novel genes under selection, providing new insights into how mosquito populations respond to insecticides and other environmental stressors.

STRUCTURAL VARIATION AND RAPID ADAPTATION

Global distribution of a deletion in an exciting insecticide resistance gene. Manuscript coming soon!

Most population genomic studies focus on single-nucleotide variants, but large structural changes in the genome can have substantial effects on gene function and likely play an important role in adaptation. I use long-read sequencing to investigate structural variation that would otherwise be difficult to detect using conventional short-read approaches.

In wild populations of the disease-vector mosquito Culex quinquefasciatus, I identified thousands of structural variants, including variants overlapping genes involved in insecticide resistance and environmental sensing. Several large-effect variants also coincided with regions showing signatures of selective sweeps, highlighting their potential role as targets of natural selection. This work demonstrates that structural variation represents an important and underappreciated source of adaptive variation and highlights the value of long-read sequencing for understanding how organisms respond to strong selective pressures. 

SPECIATION IN THE OCEAN

There are two species of Mnemiopsis along the US Atlantic coast, M. leidyi (population samples north of Roanoke Island) and M. gardeni (population samples south of Roanoke Island).

The open ocean presents a fascinating system with which to study speciation. In the ocean, organisms can disperse over vast distances and geographic barriers to gene flow are seemingly limited compared to those in terrestrial systems. Further, much of what we know about speciation comes from terrestrial and freshwater systems and if we want to fully understand the genomic mechanisms that drive speciation, we have to broaden the species that we study to include organisms with diverse life history traits. 

An example of a project that I have worked on within this theme includes investigating speciation and adaptation within the holopelagic ctenophore, Mnemiopsis. In this work, I provide genomic evidence for two distinct Mnemiopsis species along the US Atlantic coast, M. leidyi and M. gardeni. I identified substantial genome-wide divergence, genomic rearrangements, copy-number variation, and genes under selection that may contribute to environmental adaptation. By combining population genomics with demographic and comparative genomic approaches, this work connects climatic history and oceanographic processes with the genomic evolution of species.