Outline: many topics, few covered today.


1. Evolutionary perspective...

2. Software development for evolutionary genomics research.

3. Methods for inferring phylogenetic networks.

4. Modeling the evolutionary effects of species interactions.

5. Modeling the accumulation of speciation genes.

What is an Evolutionary Biologist?

Modern research in ecology and evolution requires a diverse skill set, from organismal biology, to genomics, statistics, and computational biology. My research and teaching centers on integrating these skills.


How can we most accurately reconstruct the
evolutionary history of organisms from their genomes?

How can we use genomics to reconstruct historical
ecological interactions among species?

Genomics and Botany


And it's an exciting time for this! Genomic technologies are revolutionizing the study of ecology and evolution.

Genomics and Botany


And it's an exciting time for this! Genomic technologies are revolutionizing the study of ecology and evolution.

Genomics and Botany


And it's an exciting time for this! Genomic technologies are revolutionizing the study of ecology and evolution.

Genomics and Botany


And it's an exciting time for this! Genomic technologies are revolutionizing the study of ecology and evolution.

How can we most accurately reconstruct the
evolutionary history of organisms from their genomes?

Inferring phylogeny from genome-scale data

The 'SVDquartets' framework


"invariants" are SNP patterns that occur equally (sensu word embedding).

Order SNP counts into 3 matrices for each quartet of samples.

One matrix contains more patterns matching the tree (e.g., AABB)

Two matrices contain patterns discordant with the tree (e.g., BABA, ABBA)

The matrix matching the tree aligns along rows/cols (invariants align).

SVD measures matrix filling (rank), and quickly finds best tree.

Shortcoming...

SVDquartets discards the other two subtrees which contain information relevant to inferring admixture, although we know the relative frequencies of these subtrees is informative about introgression (e.g., ABBA-BABA imbalance) (Durand et al. 2011)

In addition...

SVDquartets only examines individual quartets at a time Because quartets are not independent of each other, introgression of one may affect multiple other quartets (Eaton et al. 2012, 2015). Ideally, all taxa, or all quartets, would be examined simultaneously.

Joint inference of all quartet matrices for network inference


SVDquartets reduces SNP matrices to categorical results


SVDquartets reduces SNP matrices to categorical results


SVDquartets reduces SNP matrices to categorical results


SVDquartets reduces SNP matrices to categorical results


SVDquartets reduces SNP matrices to categorical results


Stacked count matrices

Unique fingerprint for different admixture scenarios

Simulate data and train ML model for each admixture scenario

ExtraTrees Classifier (sklearn) accurately identifies admixture edge and directionality with little training and requiring relatively little data (tens of thousands of SNPs).

Coding, reproducibility, software development


Efficient handling of large data sets, clear and reproducible coding skills, statistical literacy, understanding of data within a specialized field (e.g. genomics).

Coding, reproducibility, development

How can we use genomics to reconstruct historical
ecological interactions among species?

Floral diversity in Pedicularis

Pedicularis L. in China

Species rich:
>600 species worldwide, approximately 300 endemic to Hengduan.
We collected >60 species from 100 locations in 2018.

Morphologically diverse:
Spectacular floral diversity and abundant homoplasy; similar forms have evolved repeatedly.

Complex history of assembly:
Mountain uplift over millions of years, glacial cycles over thousands of years, river and mountains barriers, lead to constantly shuffling communities (and species interactions).

Reproductive interference

Negative fitness consequences imposed by one organism on another by disrupting successful reproduction.

Morphological terminology

The beak of the galea directs pollen placement and pickup

Elongate styles

Elongate styles have evolved multiple times (Ree 2005) and facilitate pollen competition among species (Tong and Huang 2016).

Reproductive character displacement

Hypothesis: Differences among populations (within species) are a result of interspecific interactions driving character displacement in local communities.

Thanks you




*EEEB GU4050,
PROGRAMMING AND DATA SCIENCE SKILLS FOR BIOLOGY 3 pts.


*EEEB GU4055,
PRINCIPLES AND APPLICATIONS OF MODERN
DNA SEQUENCING TECHNOLOGY 3 pts.