Dollo-CDP: a polynomial-time algorithm for the clade-constrained large Dollo parsimony problem

Table 2 This table shows a comparison of the tree produced by Dollo-CDP-fast and a best-scoring tree found by the slow heuristic search (SlowH)

# of	Same	Better	Worse	Dollo-CDP-fast	SlowH
characters	# reps	# reps (\(\Delta\) score)	# reps (\(\Delta\) score)	Runtime	Runtime
50 taxa
500	25	0 (NA)	0 (NA)	2	8
1000	25	0 (NA)	0 (NA)	2	11
5000	25	0 (NA)	0 (NA)	3	27
10000	24	0 (NA)	1 (14)	4	40
50000	25	0 (NA)	0 (NA)	9	114
100 taxa
500	24	1 (1)	0 (NA)	27	28
1000	24	1 (3)	0 (NA)	23	75
5000	25	0 (NA)	0 (NA)	10	206
10000	24	0 (NA)	1 (7)	13	333
50000	23	0 (NA)	2 (56)	33	935
200 taxa
500	19	5 (1)	1 (1)	131	132
1000	25	0 (NA)	0 (NA)	183	174
5000	17	0 (NA)	8 (3)	77	2358
10000	19	0 (NA)	6 (6)	103	2362
50000	17	0 (NA)	8 (24)	145	5638

We report the number of replicates for which Dollo-CDP-fast is same, better, or worse than SlowH in terms of Dollo criterion score. We also report the absolute difference in scores for the trees estimated by the two methods, averaged over the better and worse cases, respectively (rounded to the nearest integer). Lastly, we report the runtime (in seconds), averaged across all 25 replicates (rounded to nearest integer). Recall that the runtime of Dollo-CDP-fast includes the time to run FastH, the output of which is used to build constraints

ISSN: 1748-7188