Figure 4

Agreement between model parameters from the six replicates. (a) The Naef model scaled affinity parameters. They show good consistency, except for the behavior of guanine near the probe attachment point (nucleotide position 25). (b) Zhang model scaled 'binding energy' parameters for each of the three control samples (red circles, triangles and crosses) and for each of the three spike-in samples (green circles, triangles and crosses) for each di-nucleotide pair. In addition, the average over the six samples is indicated with black circles and the average over the two sets of energy parameters distributed for seven chip types distributed with Perfect Match [26] is indicated with black triangles and squares. The Zhang energy parameters are not as consistent as the Naef parameters, especially for AG and GA di-nucleotides. (c) Zhang's weights parameters for the six experiments (red), their mean (black line) and the average of the weights for the seven sets of weights (for non-specific and specific binding) distributed with the PDNN program (dotted lines). The parameters refined here show a clear difference from the averages over the two sets of weights distributed with PDNN. In all cases, these weights confirm the importance of the central part of the probe.