The likelihood at all sites with pattern x in a dataset of size N was then calculated as
A 6-taxa dataset of length 1000 bp and a 15-taxa dataset of length 3000 bp were both generated using the program SeqGen of Rambaut and Grassly (1997).
To corroborate this analysis, we numerically estimated the likelihood surfaces of expected datasets of increasing length, each composed of character state patterns at their expected frequencies given the Felsenstein-zone star tree in Fig. 1a.
The '—' value means that the executions with datasets of that length did not finish within three days, and were aborted because it makes no sense to compare software tools with such long execution times.
Two multi-locus functional QTL analysis approaches were proposed and their performances were compared on trait datasets comprising 2 to 9 time points, 91 to 455 individual tree measurements and genotype datasets of amplified length polymorphisms (AFLP), and single nucleotide polymorphism (SNP) markers.
The difference between average ML-estimated rates for different depths was also <2-fold for simulated datasets of branch lengths of ≤1 substitution/site; however, although it was not significant in any case, the difference in averaged estimated rates was considerably higher for longer branch lengths (Table 1).
We analyzed the FANTOM3 and H-Invitational datasets of full-length mouse and human cDNAs, respectively, and found that in transcription units with multiple start sites, the inclusion of at least 15% and possibly up to 30% of the 'cassette' exons correlates with the use of specific transcription start sites (TSS).
Next, for each subject and condition, the filtered datasets of varying lengths were segmented into contiguous 1.3 s long epochs, where the number of artifact free samples ranged from 381 to 1,074 for the 20-min EC conditions (BL, MID, and POST) and from 100 to 455 for the 10-min long EO conditions.
These datasets contain reads of length between 150 to 350 bases.
