New Approaches for Phylogenetic Species Tree Estimation

on YouTube ↗32m

" summary

Three concrete claims: quartet theory gives statistically consistent species-tree inference under ILS/GDL/HGT and is implemented in ASTRAL/ASTRAL‑Pro which beats concatenation on high discordance; a divide‑and‑conquer pipeline lets maximum‑likelihood tools (RAxML/IQ‑TREE) scale from thousands to tens of thousands of taxa, replacing multi‑hundred‑CPU‑year concatenation runs with minutes to hours; and co‑estimation (BAli‑phy) wins on simulated alignments but fails on structural benchmarks, posing a real alignment model vs benchmark conundrum.

AI-written summary · about

" chapters (23)

1. 0:00Intro
2. 0:54Phylogenomic pipeline
3. 1:411KP: Thousand Transcriptome Project
4. 2:14Avian Phylogenomics Project
5. 3:30Large datasets are difficult
6. 4:22This talk 1. Estimating species trees from gene trees ("easy")
7. 5:29Phylogeny Estimation
8. 6:16Gene tree discordance
9. 7:41Traditional approach: concatenation
10. 8:30Summary Method Protocol
11. 13:14Summary and two key ideas for ILS-based species tree estimation
12. 13:50Gene Family Trees
13. 14:19Species tree estimation under GDL
14. 18:17What about HGT?
15. 19:08Gene trees - Species trees, using Quartet Trees
16. 19:36Part II: Large-scale gene tree estimation
17. 20:05DNA Sequence Evolution (Idealized)
18. 20:50Maximum likelihood for gene tree estimation
19. 22:59Part III: Multiple Sequence Alignment
20. 23:12Multiple Sequence Alignment (MSA): a scientific grand challenge!
21. 24:46Simulation Studies
22. 27:11Statistical Alignment
23. 28:51What is going on? Most likely not an issue of failure of the MCMC analyses to converge (48 hours, 32 processors, 30 sequences).