Table 3. Comparison of the KFV algorithm with other methods for motif retrieval using Dataset-1.
Accuracy | |||
Method | Non-ZF PFMs(71) | ZF PFMs (25) | Total (96) |
KFV (kâ=â4, cosine) | 0.915 | 0.600 | 0.833 |
STAMP (PCC) | 0.887 | 0.600 | 0.813 |
STAMP (SSD) | 0.859 | 0.560 | 0.781 |
STAMP (AKL) | 0.831 | 0.520 | 0.750 |
STAMP (ALLR-LL) | 0.859 | 0.400 | 0.740 |
STAMP (pCS) | 0.761 | 0.560 | 0.708 |
STAMP (ALLR) | 0.775 | 0.400 | 0.677 |
MOSTA (Smax) | 0.915 | 0.440 | 0.792 |
MOSTA (Ssum) | 0.817 | 0.560 | 0.750 |
The results are shown separately for the zinc-finger and non zinc-finger families. The values in bold indicate the highest accuracy achieved for each category. In parentheses beside each method are the primary parameter settings (column comparison metric for STAMP or similarity measure score for MoSta). The accuracy for STAMP using different column comparison metrics were taken from [8], in which the evaluation was performed using the optimal alignment strategies and gap scores on the same dataset. For MoSta, a GC content of 0.5 and the balanced threshold were used.