Sophie Jean Walton

BAPG 2025 was a blast. Looking forward to seeing folks in Davis in Spring 2026! bapg2025.github.io/bapg2025stan...

Super excited about our schedule for BAPG at stanford on Dec 6. bapg2025.github.io/bapg2025stan... Amazing talks, a fabulous keynote, a lively poster session. A brilliant and interactive community. What’s not to love? 1/2 @sophiejwalton.bsky.social

Registration for BAPG 2025, Stanford Dec 6 2025 Talk submissions are now closed, but poster submissions are still open.

Please note that we already have 130+ people signed up and we will limit the sign up at 160. Don’t dawdle If you want to attend. Registration is free but required. forms.gle/M5BSyAaeUggn...

We have 115 people signed up already for BAPG! Great representation across career stages too. If you want to attend don't forget to sign up. It is free but required. @sophiejwalton.bsky.social and I are looking through the wonderful talk submissions and will announce the talks shortly. Stay tuned!

High-resolution mapping of a rapidly evolving complex trait reveals genotype-phenotype stability and an unpredictable genetic architecture of adaptation The extent to which adaptation can be predicted, particularly for traits with complex genetic bases, is unknown. Here, we leveraged a model complex trait, model species, and high-powered longitudinal ...

Thrilled to finally share the magnum opus of my PhD that focuses on the genetic basis of evolutionary change! Specifically, we know we can map the genetic basis of a trait, but can we tell which genes will underlie the trait shift when it evolves? doi.org/10.1101/2025...

Dynamics of dN/dS within recombining bacterial populations The ratio of nonsynonymous to synonymous substitutions (dN/dS) encodes important information about the selection pressures acting on protein-coding genes. In bacterial populations, dN/dS often decline...

The first is from former PhD student Zhiru Liu @zzzhiru.bsky.social (now in @bengrbm.bsky.social's group @ MSK) examining the long-term patterns of selective constraint – measured by the classical ratio of nonsynonymous to synonymous mutations (dN/dS) – within recombining populations of bacteria.

don't forget to register for BAPG at Stanford on Dec 6! @petrovadmitri.bsky.social and CEHG are hosting. talk submissions close on Nov 16 - bapg2025.github.io/bapg2025stan...

docs.google.com/forms/d/e/1F...

Beyond excited to share my PhD work thus far, now up as a preprint! We found that a transposable element insertion is responsible for the recent evolution of an novel color trait. Feeling thankful to everyone who has helped in this project and thrilled to continue learning about "sparkle"!

i will never stop laughing

Thank you!

And of course thank you to @benjaminhgood.bsky.social and @petrovadmitri.bsky.social for mentoring me through this adventure. n/n

Thanks to all my amazing collaborators! @ksxue.bsky.social , Jonas and Richa were instrumental to setting up this project! Also, huge shout out to Qing, @hgellert.bsky.social and Chih-Fu who contributed to the analysis and experiments. (20/n)

This suggests that conspecific strains can behave more like ecological species when competing within larger communities, even when their genomes appear to evolve as a single biological species. (19/n)

Together, our results illustrate that selection on conspecific strains in large communities is strong, but attenuates as communities stabilize in a manner reminiscent of negative frequency dependent selection. (18/n)

There are additional analyses in the paper that I didn’t cover here, including looking at how selection shifts across nutrient environments! (17/n)

Alt text: Data showing community cohesion arranged as a table. Rows are collision types, columns are species. Data shows that winners from both hosts are present in most collisions.

We also examined whether strain-level competition outcomes were correlated across species. Interestingly, we found little evidence for strain-level cohesion during coalescence, despite prior co-selection during community assembly and within each host. (16/n)

Predicting the first steps of evolution in randomly assembled communities - Nature Communications Evolution often occurs within complex communities, but the way this context controls the rate and impact of evolution is poorly understood. The authors employ a model of resource competition to study ...

However, these findings are consistent with recent ecological theory by my amazing labmate @johndmcenany.bsky.social which suggested that conspecific strains may coexist in saturated communities even when new strains displace other species. (15/n) www.nature.com/articles/s41...

Alt text: Number of species over passages in community mixtures and control communities. Number of species in both cases equilibrates to ~50 species.

Our observations of strain-level coexistence were surprising because community mixtures did not have elevated diversity compared to autologous controls! (14/n)

Since the abiotic environment was held constant, we attributed these shifts in selection to biotic interactions within the community. This suggests that conspecific strains can coexist through niche partitioning in species rich communities. (13/n)

Alt text: Two panels. Top- An example of time varying selection for an f. plautii competition. Under constant selection, the one strain is expected to go extinct. However, selection attenuates and the strains continue to coexist. Bottom- Correlation between predicted strain diversity (f x (1-f)) and actual strain diversity. In most cases, observed strain diversity exceeds predicted strain diversity.

However, selection shifted at later generations in many competitions! Moreover, these shifts also tended to be biased towards maintaining coexistence. (12/n)

Alt text: Fig panel showing the distribution of early phase fitness (between passages 1 and 2) across all pairwise competitions.

We then estimated the time-varying relative fitnesses of conspecific strains. We found that selection during earlier generations was quite strong and predicted that one strain would dominate in most competitions. (11/n)

Alt text: Fig panel showing Bacteroides dorei strain dynamics across 4 replicate competitions. Annotations show that sampling error is small and the communities undergo large shifts in frequency. Dynamics across replicates are correlated. Empirical cumulative distributions are also shown of the distribution of frequency shifts due to sampling error, between replicate collisions and between timepoint 0 and 7.

We found that the dynamics of competing strains were characterized by large and deterministic shifts. These shifts exceeded what could be generated by measurement noise or drift, demonstrating that these dynamics were driven by strong selection. (10/n)

Alt text: Fig describing community coalescence. Two stabilized communities derived from different hosts are mixed in equal ratio to create a third community. The community that results from collision is passaged for 7 passages. Additional panel showing Marker SNVs for Escherichia coli is shown. Groups of correlated SNVs show that the E. coli strains from different hosts undergo strong selection.

We then monitored the dynamics of competing conspecific strains from each host through the course of the coalescence through shotgun metagenomics. This approach allowed us to follow pairwise strain competitions across many species in parallel! (9/n)

Alt text: cartoon describing how communities from fecal samples are derived. Stool from a healthy human donor is grown in rich media (mBHI or mGAM) w/ passages every 48 hrs. Communities are sequenced via metagenomic sequencing. Relative abundance bar plots are shown of community species composition over 5 passages (38 generations). Communities have many prominent gut families, including bacteroidaceae. Number of species over time is also shown. Communities stabilize to ~50 species.

We collected fecal samples from donors in the Bay Area and assembled in vitro gut communities from these samples in anaerobic conditions. Communities from different hosts shared some species, but harbored diverged strains of those species. (8/n)

Alt text: Fig panel from Li et al 2016 showing that donor and recipient strains in an FMT can coexist. Results are displayed as a table with species as rows and FMT recipients as columns. The outcomes of the FMT vary at both a species and subject level.

Microbial community coalescence experiments have been used to gain a lot of intuition on species-level selection. We were also inspired by strain level data that had been obtained from FMTs in therapeutic contexts (fig from DOI: 10.1126/science.aad8852). (7/n)

Alt text: Cartoon of community coalescence. Communities from two hosts are combined to create a third community.

It’s hard to distinguish between these models in natural microbiomes, so turned to lab systems to get some intuition. Specifically, we performed whole community competitions, or community coalescence, with in vitro gut communities derived from different human donors. (6/n)

Alt text: Muller plot of three species fluctuating in abundance over time. A new strain of one of the species invades from a global population and competes with the old strain.

Others have proposed that conspecific strains may compete as ‘ecological species’ and coexist via niche partitioning because we see lots of examples of multiple conspecific strains persisting in the same microbiome. (5/n)

However, some ecology theory suggests that diverse communities could select for suites of strains with small fitness differences within a community. DOI: 10.1016/j.tree.2006.08.003(4/n)

Microbiologists have shown us that conspecific strains can have large differences in important traits, such as carbohydrate metabolism. This could mean that conspecific strains have large fitness differences in any local environment (3/n).

Alt text: Cartoon phylogeny of 4 human hosts and their microbial strains. Arrows indicate that these strains are separate by ~100kya of evolution. ~10k SNVs separate the microbial strains.

Complex microbial communities, like the human gut microbiome, can harbor extensive strain-level diversity, with strains of the same species from different hosts differing by ~10k mutations. However, the fitness consequences of this diversity in community contexts are not well understood. (2/n)