SPEEDEMON number of topologies in the posterior distribution differs from densitree

[Diego Caraballo]

Hi all, I am having trouble interpreting the output from a Speedemon analysis. The output generated by ClusterTreeSetAnalyser states that there are 5 different topologies, and the 95% most credible set contains 3 of them:

The 95 precent credible set consists of 3 out of 5 topologies

support #taxa topology
45.32% 5 ((((andeanus + groupD + groupE,hortensis),colcaensis),dichrous + lunatus),leiotettix)
29.47% 6 (((((andeanus + groupD,groupE),hortensis),colcaensis),dichrous + lunatus),leiotettix)
22.46% 7 (((((andeanus + groupD,groupE),hortensis),colcaensis),(dichrous,lunatus)),leiotettix)
2.74% 6 ((((andeanus + groupD + groupE,hortensis),colcaensis),(dichrous,lunatus)),leiotettix)
=== 95% credible set ===
0.01% 8 ((((((andeanus,groupD),groupE),hortensis),colcaensis),(dichrous,lunatus)),leiotettix)
100% displayed

support #taxa clusters
45.32% 5 andeanus + groupD + groupE,colcaensis,dichrous + lunatus,hortensis,leiotettix
29.47% 6 andeanus + groupD,colcaensis,dichrous + lunatus,groupE,hortensis,leiotettix
22.46% 7 andeanus + groupD,colcaensis,dichrous,groupE,hortensis,leiotettix,lunatus
2.74% 6 andeanus + groupD + groupE,colcaensis,dichrous,hortensis,leiotettix,lunatus
=== 95% credible set ===
0.01% 8 andeanus,colcaensis,dichrous,groupD,groupE,hortensis,leiotettix,lunatus
100% displayed

support count cluster
100% 18133 hortensis
100% 18133 colcaensis
100% 18133 leiotettix
74.79% 13561 dichrous + lunatus
51.94% 9418 groupE
51.93% 9416 andeanus + groupD
48.06% 8715 andeanus + groupD + groupE
25.21% 4572 lunatus
25.21% 4572 dichrous
0.01% 2 andeanus
0.01% 2 groupD

support count pair
99.99% 18131 andeanus + groupD
74.79% 13561 dichrous + lunatus
48.06% 8715 andeanus + groupE
48.06% 8715 groupD + groupE

However, when I check the .trees file in Densitree, I find that there is only one tree, using the same burnin (10%). It also happens if I turn the burnin to 0, both in ClusterTreeSetAnalyser and in Densitree.

So, I don't understand why there are 3 topologies (up to 5!) in the ClusterTreeSetAnalyser while only one in Densitree.

Here is a link to the speedemon output and the BEAST/SNAPPER output files, also.

I hope you can help me in this regard.

Best wishes,

Diego

[Diego Caraballo]

Forgot to say that the epsilon value is 0.38 (in case you want to run  ClusterTreeSetAnalyser)  😊

[Jordan Douglas]

Hi Diego,

The discrepancy arises because DensiTree is reporting the taxon-topologies, while ClusterTreeSetAnalysis is reporting the cluster-topologies. For example, in topology 1, the taxa "andeanus + groupD + groupE" form a single cluster at this value of epsilon and hence they form a single tip in the topology, whereas DensiTree is agnostic to the concept of clustering and the value of epsilon so it treats them as 3 different tips. Overall, there is just a single topology post-burnin if you do not look at node heights or epsilon, and several possible topologies once you start clumping taxa together.

Hope this helps,

Jordan

[Diego Caraballo]

Thank you, Jordan. That was exactly was I suspected but I wanted to have an authorized confirmation.  ClusterTreeSetAnalysis is cluster aware, and the alternative topologies are shrinked versions of the topology obtained in the standard snapper analysis, according to the selected value for epsilon.

Thank you very much!

Best,

Diego