Table 2. Comparison of several methods using the LSU rRNA sequences.
| Method | CPU time (s) | Sum-of-pairs score | Column score |
|---|---|---|---|
| 72 sequences × 1305–5183 sites | |||
| CLW18d | 1998 | 0.692 | – |
| CLW18q | 600.2 | 0.597 | – |
| NW-AP-2 | 197.0 | 0.796 | – |
| NW-NS-2 | 205.2 | 0.770 | – |
| FFT-NS-2 | 73.39 | 0.769 | – |
| FFT-NS-i | 251.8 | 0.781 | – |
| 59 sequences × 2810–5183 sites | |||
| T-COFFEE | 35 860 | 0.806 | 0.559 |
| CLW18d | 1523 | 0.754 | 0.411 |
| CLW18q | 395.6 | 0.643 | 0.315 |
| NW-AP-2 | 153.7 | 0.823 | 0.482 |
| NW-NS-2 | 159.8 | 0.793 | 0.463 |
| FFT-NS-2 | 51.09 | 0.794 | 0.468 |
| FFT-NS-i | 181.7 | 0.817 | 0.552 |
The CPU time and the sum-of-pairs and column scores (8) of each method are shown for the full LSU data set (72 sequences) retrieved from RDP-II (33), which includes some fragment sequences, and a subset composed of 59 sequences without fragment sequences. In the case of the full LSU data set, the column score cannot be calculated because of fragment sequences, and the test of T-COFFEE is aborted due to a memory shortage. The default parameters were used for all methods, except for CLUSTALW, in which default setting (CLW18d) and ‘quicktree’ option (CLW18q) were examined. All calculations were performed on a Linux operating system (Intel Xeon 2 GHz with 4 GB of memory). The gcc compiler (version 2.96) was used with the optimization option ‘-O3’.