genome-wide scans for variants under selection

applying machine learning and computer vision methods to images of genetic data

This was a project as part of my dissertation thesis

Building on our work leveraging ancestry patterns to detect and characterize signatures of rapid adaptation in admixed populations, I am currently working on developing methods to improve our ability to study the evolutionary history of admixed populations.

I applied machine learning and computer vision methods to images of admixed chromosomes painted by genetic ancestry. Typical population genetics methods rely on one or more summary statistics. This image-based appraoch allows us to leverage all the information provided by genetic ancestry.

Simulated admixed chromosomes are painted by ancestry. We apply computer vision methods like object detection or image segmentation to localize variants of interest along the chromosome.

In addition to optimizing these methods for this type of data, I also identified the timescales, genomic features, and demographic/selective histories that are best suited for this ancestry-based detection of genetic variants of interest through simulations. This work was published in 2023 in MBE.