The median motif width was 17 bp.
Tables 8 and 10 present the mean error in motif width based on both data collections.
Although the mean error in motif width for models predicted by MCOIN appears to decrease with decreasing motif conservation in the E. coli data collection, this is explained by the small number of datasets tested.
Results of tests on two data collections of previously characterised prokaryotic motifs show that MCOIN outperforms the E-value of the resulting multiple alignment (currently the most widely used estimator) as a predictor of motif width, using mean absolute error and root mean squared error.
Based on tests with previously characterised prokaryotic TFBS motifs, we show that MCOIN outperforms the E-value of the resulting multiple alignment as a predictor of motif width, using mean absolute error.
This performance is assessed here through the mean absolute error (MAE) and root mean squared error (RMSE), comparing the predicted motif width to the known width.
Using realistic synthetic data and previously characterised prokaryotic data, we show that MCOIN outperforms the current most popular method (E-value of the resulting multiple alignment) as a predictor of motif width, based on mean absolute error.
We first note from Table 4 that the width predicted by MCOIN closely matches the true width in almost all cases; the error in the predicted width increases slightly as mean motif conservation is decreased.
This subtlety means that the most statistically significant motif width need not match the biologically known true motif width.
This means that, even if the true motif width were known in advance, a motif discovery algorithm is not guaranteed to discover this motif perfectly.
