Patrick McKenzie
LightNews LightNews
banner
patrickmckenzie.bsky.social
Patrick McKenzie
@patrickmckenzie.bsky.social
1.4K followers 510 following 140 posts
Postdoc in Harvard OEB -- Hopkins lab. systematics, speciation, hybridization in flowering plants. Natural history via community science. Previously Eaton lab at Columbia. Birder, Arkansan, he/him
Posts Replies Media Videos
patrickmckenzie.bsky.social
Really enjoyed it, sorry I had to run right after or I would’ve said hi!
November 8, 2025 at 2:00 PM
patrickmckenzie.bsky.social
super helpful guide!!!! and congrats on the milestone!
September 18, 2025 at 8:54 PM
patrickmckenzie.bsky.social
super grateful to colleagues and reviewers who greatly improved early versions of this manuscript. We're really proud of it.
September 11, 2025 at 7:51 PM
patrickmckenzie.bsky.social
there's value in leveraging linked genomic signals (vs. just unlinked gene trees) to learn about species history and demography. In the future we're hoping to see this plugged into species-tree inference and/or ARG inference pipelines.
September 11, 2025 at 7:45 PM
patrickmckenzie.bsky.social
We implement the math in ipcoal, and we validate against coalescent simulations. Finally, we demonstrate that there is information from observed waiting distances alone for fitting MSC parameters (Ne, divergence times, recombination rate).
MS-SMC predictions validated against coalescent simulations. (a) Results are shown for three models containing 1, 2, or 8 populations. For each model, 100K tree sequences were simulated for 10 different constant Ne values between 50K and 500K. (b) The distributions of summed edge lengths of the first genealogy in each tree sequence. (c-d) The mean frequency (black square) with which the first observed recombination event was a tree-change (c) or topology-change (d) in a simulated tree sequence, and the mean (colored circle) and 95% CI (grey fill) of the predicted probability of tree or topology-change calculated from the first embedded genealogy in each tree sequence. Probabilities are constant with respect to Ne in the single population model but vary in models with population structure (also shown with respect to species tree interval lengths in coalescent units, across the top axis). (e-f) The mean waiting distance (black square) until the first observed tree-change (e) or topology-change (f) in a simulated tree sequence, and the mean (colored circle) and 95% CI (grey fill) of predicted waiting distances calculated using the first embedded genealogy in each tree sequence.
September 11, 2025 at 7:43 PM
patrickmckenzie.bsky.social
This *explicitly* links observable turnover in local ancestry to MSC model parameters.
Figure 2. An example genealogy embedded in a species tree and the corresponding embedding table used for MS-SMC calculations. (a) A four tip species tree is composed of seven discrete population intervals separated by speciation events (Wx), each of which can be further dissected by coalescent events (tx). (b) The probability of coalescence is constant within each discrete interval and is scaled by the number of lineages (k), the effective population size (n), and the interval length (d). Under the MS-SMC, the probability that a detached lineage will re-coalesce on a specific branch of the genealogy (e.g., branch 7) is calculated using the piecewise constant probabilities from each discrete interval spanning that branch (e.g., rows 1, 6, 7, and 8 in the embedding table). Figure 4. The MS-SMC is a model of the probability of different categorical event outcomes given recombination occurring uniformly on a genealogy embedded in a parameterized MSC model (a). A recombination event can cause one of three possible recombination event types between two sequential genealogies in a genome: no-change, tree-change, or topology-change. The probability of these event types are calculated by integrating over each branch on the genealogy on which recombination can occur; which in turn is calculated by integrating over each position on a genealogy branch at which recombination can occur. (b-d) The probability that a recombination event causes a no-change, tree-change, or topology- change event, for the example genealogy embedded in a species tree, was calculated for three selected genealogy branches (2: red; 5: blue; and 7: purple), at each position along the branch where recombination could occur.
September 11, 2025 at 7:41 PM
patrickmckenzie.bsky.social
alignments are mosaics of autocorrelated genealogies... nearby bases share ancestry until recombination alters the local tree. We derived the distributions of waiting distances for tree-change and topology-change events, extending prior single-population results to arbitrary species tree models.
figure illustrating the concept of an ARG, including recombination events that create no-change, vs. tree-change, vs. topology-change
September 11, 2025 at 7:40 PM
patrickmckenzie.bsky.social
Really sorry to hear this. good thoughts your way.
August 29, 2025 at 7:58 PM