Levi Yoder Raskin

Thanks so much for sharing! To add, we don't necessarily think shape is poor data for phylogenetic inference. Instead, we think PCA is inherently a poor method for phylogenetic inference! My dissertation research focuses on Bayesian inference of phylogeny from shape data; hope to share more soon!

Principal Components Analysis fails to recover phylogenetic structure in hominins Objectives: Paleoanthropologists often utilize geometric morphometrics and principal components analysis (PCA) to interpret shape variation within the hominin fossil record. It is common practice to i...

Despite its widespread use in paleoanthropology, PCA is ill-suited and unreliable for inferring hominin evolutionary relationships from morphological data. Studies of hominin systematics that rely on PCA as a means of inferring morphological and phylogenetic affinity should be revisited. end/

Much of what we know about hominin diversity comes from studies which make phylogenetic claims from PCA. Here, we simulated datasets analogous to the hominin fossil record and found that PCA vastly underperformed, recovering topology in 0-2.4% of simulated datasets. 2/

Happy to share a new preprint, the first from my PhD research with Professor John Huelsenbeck at UC Berkeley! Here, we test whether PCA can recover hominin phylogeny from both geometric morphometric data and traditional measurements. 1/ 🧪 #PaleoSky #phylogenetics

We came across your Sys Bio paper a few days ago and are absolutely including it in our revisions!! So sorry we missed it the first time around-- finding it was really reassuring to us! I'm hoping to keep working on character analysis throughout my PhD, so hopefully more interesting results to come!

I'm interested!!

I am interested! I haven’t done any sci comm but will be posting more both here and on Twitter and trying do hominin/phylogenetics sci comm.

Awesome, thanks! I'll start with the twitter threads as a way to practice/get in a habit with it and see what works. Thank you!

Do you have any suggestions for scientists trying to reach broader audiences?

I’m a PhD student in evolutionary biology (formerly studied Paleolithic archaeology) and would love to try and reach audiences that aren’t primed for evo bio, but don’t know where to start.

Thanks!

Hi, not sure if I'm too late on this. I just swapped over from twitter and I'm hoping it would be possible to join the Science feed! I'm an Integrative Biology PhD student at UC Berkeley and I study phylogenetic methods development.

Here's my orcid for proof:
orcid.org/0009-0001-26...

🧪 #evolution #phylo #PaleoSky #paleontology
*Still learning bluesky and feeds*

The effects of trait redundancy and information content on hominin phylogenetic inference Paleoanthropological phylogenetic inference is based on characters assumed to be phylogenetically informative and mutually independent, but our understanding of which traits fulfill these requirements...

Preprint: biorxiv.org/content/10.1...
#paleoanthropology, #hominins, #humanevolution, #phylogenetics, #morphology, #parsimony, #characteranalysis, #methods, #BrynMawrCollege, #HaverfordCollege, #UCBerkeley

4) Informative traits are informative to the most parsimonious topology and not necessarily to the correct topology, which may never be known. When we say informative we mean that the most parsimonious consensus hominin topology is dependent on the inclusion of an unaltered trait

3) While Mongle et al. (2023) have expanded the dataset we used, we decided to use the 2019 matrix for easy comparison to other methodological papers (e.g., Gautney, 2023)

2) We also only measured the information content in traits and bony elements in Mongle et al. (2019)’s character matrix; our results are limited to traits in this character matrix and operationalized in this way

Limitations: 1) we inferred topologies using parsimony, as that method is still widely used in morphological phylogenetics (follow for updates on novel Bayesian approaches as my PhD progresses!)

A better understanding of missing data, informative data, and what traits are driving the cladistics seen both in our paper and many other hominin phylogenetic papers is crucial to understanding the uncertainty inherent in the record

For example, below is (left) the MP consensus topology using all taxa and traits and (right) the MP consensus topology inferred from the character matrix missing Kenyanthropus platyops. Purple diamonds indicate changes in relation to the topology including all the taxa

We also find that isolated cranial bones (e.g., isolated fossil mandible) contain insufficient information to closely resolve the hominin topology and demonstrate that the baseline hominin topology is potentially biased by the inclusion of dubious taxa

We find that only 21% of the characters in that matrix contribute unique phylogenetic information and at least 22 traits (and perhaps up to 76) contribute no information whatsoever to the inferred topology under a maximum parsimony framework

We use a combination of permutation of trait values and removal of traits, sets of traits composing anatomical units, and operational taxonomic units

Happy to share my first preprint from my undergraduate work at Bryn Mawr College with Professors Maja Šešelj and Bárbara Bitarello! We present new methods to assess a hominin craniodental discrete character matrix (107 traits) for phylogenetic information. (1/12)