MGDrive: Landscape Clustering (NCUR Oral Presentation) (Gillian Chu)

The idea in this sets of experiments is to understand the differences in dynamics of the spread of gene drives in spatiotemporal landscapes, and how these dynamics are affected when we increase the aggregation (clustering) level of the breeding sites.

The main goal is to determine a metric of information loss that we can use in the specific context of gene flows in mosquito populations.

For this experiment, we are defining equally-sized mosquito populations to lie in a line, at a uniform distance from each other. With this basic spatial setting, we cluster hierarchically the populations: starting from having each node in it’s own “cluster”, and going up to all the points belonging to the same cluster group.

In doing so, we increase the population size to keep that variable constant across scenarios.

![convergence.jpg]

After generating different levels of aggregation, we run a gene drive construct. In this case, we are using MCR with different configurations $[H_{fitness},B_{fitness},F_{deposition}]$ to test various fixation speeds.

After running MGDrivE, we display de results of the simulation runs to look at the qualitative differences in the systems’ responses.

This video illustrates how the code was used:

I didn’t end up getting to present this, because the conference was cancelled due to the pandemic. However, the paper will be published in the NCUR Journal, and linked to here when it becomes available.

Read more about MGDrive here. My work here was also presented as part of the UC Berkeley Center for Computational Biology Retreat in 2018 here.

[2020 March] MGDrive: Landscape Clustering (NCUR Oral Presentation) (Gillian Chu)

[2018 October] MGDrive The Original Trilogy (Computational Biology Retreat Oral Presentation) (Hector M Sanchez C., Jared Bennett, Valeri Vasquez, Sean L. Wu, Sarafina Smith, Gillian Chu, Tomas Leon, John M. Marshall)