Over the past two weeks
I’ve been attracted to poor Bootstrap scores in the large reptile tree (LRT, 1151 taxa, subset Fig. 1) reexamining data and re-scoring where necessary. The result is a tree with improved Bootstrap scores. Herewith, the eutherian (placental) mammal subset of the LRT.
will find the one node that is not resolved in this tree. Hint: the specimens lacking resolution are known from damaged skulls and a few post-cranial bones, so they can be scored for a relatively few character traits.
seeking more information for any genera listed above need only use it for a keyword in the search feature of this blog post (above).
the present tree has been improved, there is still room for improvement, probably around the weaker Bootstrap scores.
of just the basal tetrapods and lepidosauromprhs (370 taxa, 1 tree) took less than 51 seconds for a completely resolved subset of the LRT. Testing of just the archosauromorphs (781 taxa, 2 trees) took 8:45 minutes of computing time. So, 410 more taxa and one more tree take more time.
Taking it to the final step: Testing of the entire LRT (1151 taxa, 14 trees) took 1 hour 50 minutes. You can see computing time rises exponentially with increasing taxa, even with the next best thing to complete resolution.
So where did those 12 extra trees come from?
Should be from no more than 3 unresolved nodes. Here’s where PAUP fails (or becomes exhausted) with high taxon numbers:
Basalmost Synapsida (Ellioitsmithia, Apsissaurus, Aerosaurus, etc.), Lepidosauria/Sphenodotia/Marine Younginiformes/Diadectomorpha + Pareiasauria/ Caseasauria/Basal Lepidosauromorpha/ Basal Archosauromorpha/ Basal Diapsida/ 13 more little clades/15 single taxa and…Hypuronector/Vallesaurus/Megalancosaurus
So with all those problems
(way more than expected) I ran PAUP again, sans mammals and terrestrial younginiforms (including protorosaurs and archosauriforms): so…. basically all the primitive taxa were included. Result: 565 taxa, 2 trees) took 5:54 minutes with loss of resolution between (Megazostrodon + Hadrocodium) and (Brasilitherium + Kuehneotherium), three of which are skull-only taxa just outside of the deleted mammals. No other tree topology changes are recovered.
Just so you know…
it seems that PAUP does exhaust itself in large cladograms, even in a simple Heuristic search.