Figure 8

In vitro binding affinity of NF-κB's RelA and c-Rel proteins to κB sites correlates well with HMM-predicted binding occupancy probability. Gel shift assays with extracts from 293T cells transiently transfected with either CMV-hRelA (A), CMV-hc-Rel (B) or empty CMV vector as control (vector) and radiolabeled double-stranded oligonucleotide probes containing the predicted NF-κB sites derived from chicken blnk site 1 or site 2, pdcd4, itm2b, pp1e, bcap, igλ, or mip-1β, or a palindromic NF-κB DNA site as control (κB-PD). Reactions containing the κB-PD probe alone, in absence of cell extract, were loaded as control (probe). DNA/protein complexes were resolved from unbound DNA probes in native 5% polyacrylamide gels. (C) shows the sum of Kullback-Leibler (KL) divergences of the HMM-predicted occupancy probabilities of the above sequences (in the gel shift constructs) with their binding affinities in the gel shift experiments, as a function of the transition probability to the motif z. The sum of the KL divergences is minimum at z equal to 0.001 for both NF-κB proteins. (D) shows the correlation between the gel shift binding affinities of the above sequences and their occupancy probabilities predicted by the HMM at z equal to 0.001. The correlation coefficients are 0.91 and 0.92 in case of RelA and c-Rel, respectively. The dashed lines in (D) are linear least square fits.