00-introduction.Rmd

# Introduction

Dobzhansky's famous quote "Nothing in biology makes sense except in the light of
evolution" has become a cliché, but it is hard to find a statement that more
elegantly captures the essence of biological study [-@dobzhansky2013nothing]. In
a more reductionist sense, it is perhaps appropriate to state that nothing in
evolution makes sense except in the light of mutation; that is to say that
mutation in an organism's genetic code is the seed upon which evolution acts
[@lynch2016genetic]. Of course, a seed left alone will not grow; it needs the
right environment and conditions. In turn, a mutation needs a population in
order to thrive. A population can be defined as a group of interbreeding
individuals, coexisting temporally and spatially (though we will see later that
this is not always the best definition) [@milgroom1997contributions]. In a
population, if the conditions are right, a new mutation can grow in frequency
and become an established allele. The study of population genetics seeks to
understand how evolution acts upon these alleles to shape populations as they
grow, shrink, split, and adapt to their environment. Ultimately, these
populations are what give rise to species.

## Population Genetics and Clonality

The classical definition of a population is reliant upon the assumption that the
individuals within the populations are panmictic, that is that they are randomly
mating so that the individuals in the next generation will have a random
assortment of alleles inherited from their parents [@rice2002evolution;
@nielsen2013introduction; @hartl1997principles]. The assumption that sex occurs
applies to most major organismal groups [@rice2002evolution;
@heitman2012evolution]. However, sex does not occur in all organisms as many
plants, fungi, protists, bacteria, animals, and viruses are known to exclusively
reproduce clonally [@rice2002evolution; @heitman2012evolution;
@arnaud2012implications; @orive1993effective; @arnaud2007standardizing]. Clonal 
reproduction here means reproduction that results in offspring that are
genetically identical to the parents (note that this can include organisms
derived from the selfing of a haploid organism). Thus, the definition of a
population is perhaps best rephrased to simply indicate a group of individuals
of the same species^[This depends on your definition of a species, which is
outside of the scope of this dissertation and therefore best debated elsewhere.]
coexisting spatially and temporally [@milgroom1997contributions].

There are several costs and benefits associated with both clonal and sexual
reproductive strategies from an evolutionary perspective. For organisms whose
main mode of reproduction is sexual, there is the two-fold cost of outcrossing:
1) only half of the available genetic material is inherited by the offspring and
2) the loss of advantageous gene combinations due to recombination
[@rice2002evolution; @heitman2012evolution; @nieuwenhuis2016frequency]. These
costs are balanced by the benefits that come with recombining genetic material,
which allows these organisms to purge deleterious mutations. As clonally
reproducing organisms do not undergo regular periods of recombination, they
explicitly do not benefit from the previously mentioned advantage of
recombination [@heitman2012evolution; @nieuwenhuis2016frequency]. This leads to
a theory known as Muller's ratchet, which indicates that the overall fitness of
small populations of clonally reproducing organisms will decrease over time as
the number of deleterious mutations accumulate, eventually leading to extinction
[@felsenstein1974evolutionary; @loewe2006quantifying; @lynch1993mutational;
@lynch1990mutation]. Despite this, clonal organisms still have the advantage of
passing on 100% of their genetic material to the next generation.

It's easy to see the advantage of both of these strategies in the context of
genetic adaptation to the environment. Clonal populations that are well adapted
to their environment get to keep the combination of genes that allowed them to
be so well adapted [@heitman2012evolution]. This strategy provides an advantage
when resources are scarce and competition is high, but does not lend itself well
to adaptation against changing environments due to low genotypic diversity
[@milgroom1996recombination]. Conversely, while sexual populations lose the
competitive advantage of conserved beneficial allelic combinations, they make up
for it with high genotypic diversity, increasing the chances of surviving change
in the environment. As we shall see in the following sections, many microbial
pathogens have the ability to reproduce both sexually and asexually
[@tibayrenc1995population; @anderson1995clonality; @milgroom1996recombination].

The question of sexual reproduction and population structure in microbial
pathogens is important for the development of rational management strategies for
pathogens or beneficial microbes [@tibayrenc1990clonal;
@tibayrenc1995population; @milgroom1997contributions; @taylor1999evolutionary].
If one were developing an antimicrobial agent on a known set of strains, the
effectiveness of this agent will be diminished if these strains represent only a
fraction of the genotypic diversity due to sexual recombination
[@taylor1999evolutionary; @tibayrenc1990clonal; @smith1993how]. Since most
inference of population structure is based on neutral models that assume sexual
reproduction, it is important to identify whether or not the population in
question is reproducing sexually or clonally before one attempts to assess
population structure and infer evolutionary processes [@tibayrenc1990clonal].

Neutral molecular markers are commonly used for population genetic analyses.
Before molecular techniques became widely available for characterization of
pathogens or microbes, the variation within most pathogenic microorganisms was
described using phenotypic traits that reflected their pathogenicity (ability to
cause disease), virulence (severity of disease), or morphotype placing them into
different phenotypes [@milgroom1997contributions; @levin1999population]. When
considering markers for population genetic analysis, one of the most important
criteria is that these markers are heritable.

A wide variety of heritable molecular markers exist for population genetic
analysis [@tibayrenc1995population; @halkett2005tackling]. It is not necessary
to expound here on the variety of molecular markers available to us, but it is
useful to take note of the common features. Generally, genetic markers come in
two major flavors, selectively neutral and selected markers
[@milgroom1997contributions; @mcdonald1997population; @tibayrenc1995population].
Selectively neutral markers are used as independent variables that can reflect
the evolutionary history of descent. For selectively neutral markers, it is
important that they be independent and unlinked within the genome
[@mcdonald1997population].

Because neutral genetic markers vary independently within a population, these
are more appropriate than selected markers to detect sexual reproduction. The
theory behind this is simple; sexual reproduction breaks up associations between
markers due to recombination and independent, random assortment. This contrasts
with clonal reproduction, which transfers the entire genome in tact to the next
generation, preserving associations [@orive1993effective; @heitman2012evolution;
@nieuwenhuis2016frequency; @mcdonald1997population; @milgroom1997contributions;
@tibayrenc1995population]. This means that not only will there be significant
associations between alleles in a clonal population, but because of their
inability to purge mutations, we also expect to see excess heterozygosity
[@tibayrenc1990clonal; @tibayrenc1996towards].

By using these basic principles, core sets of statistical analyses have been
proposed to detect sexual reproduction by analyzing diversity within loci (for
diploid organisms), linkage among loci, and diversity of multi-locus genotypes
within the population [@tibayrenc1996towards; @arnaud2007standardizing;
@parks1993study; @smith1993how; @mcdonald1997population; @grunwald2003analysis].
None of these tests alone can give a definitive answer as to the nature of
reproduction in a given population [@tibayrenc1995population; @de2004clonal;
@nieuwenhuis2016frequency]. Researchers still need to use biological evidence,
such as the observation of sexual reproductive structures, to support the
presence of sex.

## Population Genetics Under the Lens of Plant Pathology

One of the over-arching themes in both ecology and evolution is that diversity
is important for stable population and community structure
[@magurran1988diversity; @mcdonald2002pathogen]. Agricultural systems, however,
behave very differently to natural ecosystems [@milgroom1997contributions;
@stukenbrock2008origins]. Within agricultural systems, the plants and the
landscape are homogenized, providing a steady source of food for people, but
this homogeneity combined with intensive management practices also provides
fertile ground for the evolution of plant pathogens [@stukenbrock2008origins].
McDonald identified five major factors influencing the evolution of plant
pathogens in agricultural systems: 1) mutation, 2) genetic drift, 3)
gene/genotype flow, 4) mating system, and 5) selection due to host resistance
[-@mcdonald2002pathogen].

An example of how changing one of these five factors influences plant health is
clear in the oomycete *Phytophthora infestans* (Mont.) de Bary. The causal agent
of potato late blight (and a poster child for plant pathology). *P. infestans*
has changed the very demographics of human populations across the Atlantic ocean
due to its role in the Irish potato famine [@erwin1996phytophthora]. It has a
heterothallic (no selfing) life cycle that includes both clonally and sexually
derived infectious propagules (Fig.\@ref(fig:pinflife)). The clonally derived
propagules are swimming, short-lived zoospores that can be transmitted via wind
or water, whereas the sexually derived propagules are double-walled oospores.
While *P. infestans* can survive as hyphae in infected tubers, the sexually
derived oospores can survive for years in the soil, highlighting the added risk
presented by sexual reproduction.


```{r pinflife, echo = FALSE, fig.cap = figcap2, fig.scap = figscap2}
knitr::include_graphics("figure/introduction/Pinfestans_life.png")

figcap2 <- "Life cycle of the oomycete pathogen, *Phytophthora infestans* on 
potato [@piepenbring2015infestans]. M! = meiosis, P! = plasmogamy, K! = 
karyogamy. Accessed from https://goo.gl/XO7TmS"
figcap2 <- beaverdown::render_caption(caption = figcap2, figname = "fig2", to = TO) -> figscap2
```

From the time of the Great Famine to 1980, there was only one single mating
type, A1, observed in Europe, indicating that it was reproducing in a strictly
clonal fashion [@goodwin1994panglobal]. Sexual reproduction became possible when
the second mating type, A2, was introduced to Europe [@galindo1960nature].
Because of this risk, population genetic tools and methods were needed to assess
whether or not sexual reproduction had occurred. While evidence for sexual
reproduction was found [@sujkowski1994increased], it additionally appeared that
*P. infestans* was still reproducing and spreading clonally
[@goodwin1994panglobal]. @cooke2012genome analyzed isolates from 1983 to 2008
and found a rapid shift in clonal population structure, showing a dramatic
increase in the prevalence of a particularly aggressive lineage. While sexual
reproduction was not found to be prevalent, it was only via population genetic
techniques that this question could be addressed.

## The Need For Reproducible Research

Computational methods are increasingly important for answering questions
crucial to plant pathology, and as more and more of our analyses are becoming
computational in nature, we need to ensure that these methods are
reproducible. Computational methods are crucial for population genetic analyses
in the 21st century [@excoffier2006computer]. There has been a call for
standards in the analysis of clonal organisms for the past twenty years, which
resulted in a plethora of computer programs [@tibayrenc1996towards;
@arnaud2007standardizing]. It was not uncommon for a researcher to use several
programs to answer a single research question, and each program would often
require its own custom input format and would not necessarily run on all
operating systems [@excoffier2006computer; @kamvar2014poppr]. The constant
reformatting of data, and the inability to automate series of analyses made it
not only challenging to communicate the methods, but also increased the chance
for error [@excoffier2006computer; @adamack2014popgenreport; @kamvar2014poppr; 
@goecks2010galaxy].

Biologists often think about reproducibility in terms of wet lab or bench work.
Care is taken to ensure that all methods and data are faithfully recorded in a
lab notebook, but this is not necessarily extended to computational analyses.
The concept of reproducible research in scientific computing is attempting to
take the principles of reproducibility we use in the wet lab and extend them to
our downstream analyses *in silico*. This work attempts to introduce tools for
reproducible research in clonal population genetics in an open manner. When
researchers have open tools that can be easily shared, they can spend more time
asking questions and less time formatting data.

## Overview

```{r, introduction1, echo = FALSE, fig.cap = figcap1, fig.scap = figscap1, out.width = "50%"}
knitr::include_graphics("figure/introduction/dissertation.png")

figcap1 <- "A graphical representation of the three themes governing the work
presented in this dissertation." -> figscap1
```

The work presented here offers tools designed to answer questions related to
clonal population genetics and plant pathology (Fig. \@ref(fig:introduction1)).
In this context, the term 'tool' refers to software code used to apply
mathematical models or theory to population genetic data. The merit of this work
lies within the context of reproducible science, which ensures that the
computational environment used to create results can be replicated
[@buckheit1995wavelab]. The end goal was not simply the development of the
software, but rather, the development of software for the goal of analyzing our
own data. Serendipitously, by analyzing our own data with this software, we are
able to not only demonstrate that it can be used, but also demonstrate that an
entire analysis can be conducted in an open and reproducible manner. We
demonstrate the usefulness and flexibility of these tools by using them to show
evidence for multiple introductions in the Curry County, OR outbreak of the
sudden oak death pathogen, *Phytophthora ramorum*. We additionally assess the
power of the multilocus linkage disequilibrium measure $\bar{r}_d$ to detect
clonal reproduction.

### Summary of Chapter 2

To address the lack of tools for reproducible research in clonal population
genetics, we present the software package *poppr* in the R computing language
[@R2016; @kamvar2014poppr]. Previously, tools necessary for the analysis of
clonal populations were available in several stand-alone software programs, each
requiring different data input formats. Moreover, each program had different
levels of documentation and limited support for all computing platforms. The
novelty of *poppr* was to introduce indices of multilocus genotypic diversity,
the index of association, and a fast implementation of Bruvo's genetic distance,
and clone-correction over unlimited levels of user-specified population
hierarchies [@arnaud2007standardizing; @Agapow_2001; @bruvo2004simple]. Because
this was implemented in R these analyses can be performed in a reproducible
manner on all computing platforms.

### Summary of Chapter 3

The initial implementation of *poppr* contained basic tools for analysis of
clonal populations [@kamvar2014poppr], but lacked tools for custom definitions
of multilocus genotypes and performed poorly with genomic-scale data. Chapter 3
introduces an updated and improved *poppr* version 2.0. With high throughput
sequencing (HTS) data, the amount of missing data and genotyping error
increases, and the definition of a multilocus genotype becomes unclear.
Moreover, the calculation of the index of association scales poorly with an
increase in the number of loci. To address these limitations, we improved
*poppr* with new functionalities to define multilocus genotypes based on genetic
distance and calculate the index of association over random samples or windows
of SNP loci.



### Summary of Chapter 4

A newly-emerged disease of oak---called Sudden Oak Death---spread from
California to the Southwest corner of Oregon in 2001. Because of intense
management strategies, the epidemic was largely contained to Curry county for
the next 15 years. In 2011, an isolated patch of disease appeared in Cape
Sebastian, 12 miles from the nearest infected site. With microsatellite
genotyping performed across 2 labs and 15 years, we sought to describe the
spread of the epidemic in a population genetic context and ask the question of
whether or not there was evidence for more than one introduction event. This
work provided evidence supporting at least two introductions to Curry county
forests. All the analyses were performed in an open-source and reproducible
manner using R.

### Summary of Chapter 5

The index of association is a measure of multilocus linkage disequilibrium, that
is, a correlation coefficient across multiple loci. In sexual populations, loci
are randomly assorting due to recombination, resulting in a near-zero value of
the index of association. In clonal populations, recombination is non-existent,
meaning that loci are passed from parent to offspring in a non-independent
fashion, resulting in a significantly non-zero value of the index of
association. De Meeûs & Balloux [-@de2004clonal] demonstrated that this index
shows high variance with low levels of sexual reproduction, but due to
limitations in software, they were not able to perform power analyses. We used
*poppr* to investigate the power of the index of association to detect sexual
reproduction in simulated data sets generated with microsatellite and genomic
markers. This chapter provides novel insights on the power, sensitivity, and
scope of the index of association.