Jekyll2024-02-05T19:40:22+00:00https://elinck.org/feed.xmlLinck LabMontana State UniversityNew preprint : Geographic sampling and species limits in “Western” Empidonax2019-02-04T00:00:00+00:002019-02-04T00:00:00+00:00https://elinck.org/empidonax-preprint<p>Back in December, my coauthors and I posted a preprint of our study of species limits in the tricky (some might say “infamous”)
<em>Empidonax difficilis</em> / <em>Empidonax occidentalis</em> group of western North American flycatchers. Like all <em>Empidonax</em>, the Pacific-slope
and Cordilleran Flycatchers look <em>extremely</em> similar. In fact, they used to be treated as a single species, the “Western Flycatcher,”
before a series of papers by Ned Johnson led to their split based on phenotypic and genetic data. However, evidence of ample hybridization
in contact and the general overwhelming similarity of these species has meant there have always been a few skeptics who’d rather have left
them lumped. Using genome-wide DNA sequence data and 300-plus (!) samples, we asked how including populations of <em>E. occidentalis</em> south of the
US / Mexico border changed the story. Were Mexican birds simply a southern partition of the Rocky Mountain deme? Or were we overlooking
hidden diversity?</p>
<p><img src="/images/empids.jpg" alt="" /></p>
<p>What we found provides a cautionary tale about attempting to delimit species and infer evolutionary relationships with
biased or incomplete geographic sampling. As it turns out, a resident population in the Sierra Madre del Sur is a highly distinct and previously
unrecognized lineage. Meanwhile, the rest of <em>E. occidentalis</em> is what phylogeneticists call “paraphyletic,” meaning some populations are more closely
related to the Pacific-slope Flycatcher than other their own members of their own species! These results shake up the taxonomy of the group,
but also have broader implications for what we consider species, because there’s still pretty good evidence that despite their mixed ancestry geographic
groups form coherent populations. You can our thoughts on these issues in the paper, <a href="https://doi.org/10.1101/491688">available here</a>.</p>
<p><a href="./">back</a></p>Back in December, my coauthors and I posted a preprint of our study of species limits in the tricky (some might say “infamous”) Empidonax difficilis / Empidonax occidentalis group of western North American flycatchers. Like all Empidonax, the Pacific-slope and Cordilleran Flycatchers look extremely similar. In fact, they used to be treated as a single species, the “Western Flycatcher,” before a series of papers by Ned Johnson led to their split based on phenotypic and genetic data. However, evidence of ample hybridization in contact and the general overwhelming similarity of these species has meant there have always been a few skeptics who’d rather have left them lumped. Using genome-wide DNA sequence data and 300-plus (!) samples, we asked how including populations of E. occidentalis south of the US / Mexico border changed the story. Were Mexican birds simply a southern partition of the Rocky Mountain deme? Or were we overlooking hidden diversity?New preprint : MAF cutoffs and inferring population structure2017-09-18T00:00:00+00:002017-09-18T00:00:00+00:00https://elinck.org/maf-preprint<p>Inferring population structure – the subdivision of a species into groups of individuals
interbreeding with each other at a higher frequency than expected by chance – is a fundamental
goal of population genetics. Beyond its obvious relevance to systematics, taxonomy, and conservation,
understanding patterns of population structure is crucial for a range of applications, from
detecting selection and migration to identifying genetic variants associated with specific traits in
genome-wide association studies (the <a href="http://www.biorxiv.org/content/early/2017/09/07/185629">red-hot “GWAS” trend</a>).
While on the one hand, the wealth of data resulting from the advent of next generation sequencing
technology has proved a boon for describing subdivision in nonmodel organisms, it has also
provided challenges. For example, many computational methods for detecting structure were developed prior to the genomics
era and may not be suited to the unique characteristics of large SNP datasets. Of these,
Pritchard et al.’s <a href="https://web.stanford.edu/group/pritchardlab/structure.html"><code class="language-plaintext highlighter-rouge">structure</code></a> is the most widely cited,
and the basis for a raft of other methods that feature an underlying generative model and explicitly
estimate a suite of population genetic parameters.</p>
<p><img src="/images/structure.png" alt="" />
<em>Figure 2 from our preprint, showing inferred (A) population discrimination and (B) admixture across different MAF levels.</em></p>
<p>Unfortunately, at least in our lab’s experience, these model-based approaches frequently fail with modern
datasets produced by methods like RADseq. In particular, we’ve observed a pattern where all
individuals have the great majority of their ancestry assigned to a single population, producing
<a href="https://twitter.com/ethanblinck/status/908346882107228160">a “smear” pattern</a>
with the remaining small percentage assigned to informative clustering results.
Minor allele frequency cutoffs (or selecting parsimony informative sites alone) appeared to
partially help – perhaps reflecting theoretical predictions that different frequency
classes of alleles reflect different evolutionary histories and levels of subdivision. But we lacked
a real rationale for applying these filters while processing our data, and lacked an understanding of
what different levels of filtering actually did.</p>
<p>So to address these issues, C.J. and I used both simulated
and empirical (<a href="https://www.allaboutbirds.org/guide/PHOTO/LARGE/gc_kinglet_stephenparsons.jpg">Golden-crowned Kinglets!</a>)
SNP datasets to explore how MAF choice affected our ability to recover
known population limits. We looked at both programs like <code class="language-plaintext highlighter-rouge">structure</code> and non-model-based approaches
like k-means clustering with principal components. We found the former class of methods are indeed very
sensitive, but the latter class are fairly robust, speculated on why, and made a few recommendations
for best practices. You can read our new preprint <a href="http://www.biorxiv.org/content/early/2017/09/14/188623?rss=1">here</a> –
we’re hoping to submit it soon, but would love feedback in the mean time.</p>
<p><a href="./">back</a></p>Inferring population structure – the subdivision of a species into groups of individuals interbreeding with each other at a higher frequency than expected by chance – is a fundamental goal of population genetics. Beyond its obvious relevance to systematics, taxonomy, and conservation, understanding patterns of population structure is crucial for a range of applications, from detecting selection and migration to identifying genetic variants associated with specific traits in genome-wide association studies (the red-hot “GWAS” trend). While on the one hand, the wealth of data resulting from the advent of next generation sequencing technology has proved a boon for describing subdivision in nonmodel organisms, it has also provided challenges. For example, many computational methods for detecting structure were developed prior to the genomics era and may not be suited to the unique characteristics of large SNP datasets. Of these, Pritchard et al.’s structure is the most widely cited, and the basis for a raft of other methods that feature an underlying generative model and explicitly estimate a suite of population genetic parameters.Blog resurrection2017-09-17T00:00:00+00:002017-09-17T00:00:00+00:00https://elinck.org/new-blog<p><img src="/images/stuart.jpg" alt="" />
<em>Brendan skiing Ulrich’s Couloir off the summit of Mt. Stuart this May.</em></p>
<p>It’s been over a year since I last posted at <a href="http://beyondtheranges.wordpress.com.">Beyond the Ranges</a>.
I’m a little sad about this. My old blog was a labor of love, a reminder that I love taking photographs,
and the archive of a half decade of adventure between the end of college and the present day. There are a number of reasons why I couldn’t keep it up, though.
Writing for <em>The Stranger</em> and my subsequent increase in freelancing activity certainly cut in to my desire
to pen 3000-plus carefully considered words about sliding on snow.
My own interest in reading about running and skiing has
decreased lately, and blogs in general appear to be a thing of the past.
Most importantly, I needed to start a real professional website (this one), which became my primary home
on the internet, but <em>BTR</em> proved a bit too big and clunky to import from WordPress.</p>
<p>I can’t pretend I haven’t missed it, though, and I always intended to start writing casually again. Hopefully I’ll
be able to keep it more succinct and more fun, for longer. Expect fewer detailed race reports,
a more curated selection of photos, and more biology. I’ll try and get a comments section up soon – thanks to everyone
who’s stopped by over the years.</p>
<p><a href="./">back</a></p>Brendan skiing Ulrich’s Couloir off the summit of Mt. Stuart this May. It’s been over a year since I last posted at Beyond the Ranges. I’m a little sad about this. My old blog was a labor of love, a reminder that I love taking photographs, and the archive of a half decade of adventure between the end of college and the present day. There are a number of reasons why I couldn’t keep it up, though. Writing for The Stranger and my subsequent increase in freelancing activity certainly cut in to my desire to pen 3000-plus carefully considered words about sliding on snow. My own interest in reading about running and skiing has decreased lately, and blogs in general appear to be a thing of the past. Most importantly, I needed to start a real professional website (this one), which became my primary home on the internet, but BTR proved a bit too big and clunky to import from WordPress.