The agreement between the predicted and native complex was quantified using the standard rmsd superposing representing the protein by Ca atoms and the DNA by C1' nucleotide atoms. However inspection of the results suggested that this did not provide a good guide to the quality of the prediction. Instead the percentage of correct contacts in the interface was used. First, the protein-nucleotide pairs in the native interface were defined as those pairs that had at least one non-hydrogen protein-DNA atom-atom contact of <5A. For these pairs, the Ca-C1' distances were measured in the native and the predicted complexes and if the difference in distance was <4A, then a correct contact was assigned to the predicted structure. A "good" prediction was taken if the percentage correct contacts compared to the total number of interface contacts (%CC) was >65%.
Table 8.3 gives the rank of the first "good" prediction in lists ordered by the scoring function after the two types of distance filters ranking by shape complementarity and by the empirical scoring scheme. Out of the eight systems only for MET did the approach fail to generate a "good" solution. After filter 2, shape complementarity gave 4 solutions with a rank in the top 5. When ranking by the empirical score, four complexes were in the top 5 rank after filter 1 or after filter 2. However other complexes were below the top 100 solutions. At present, the empirical score would be useful to generate a limited list (say 5) of suggestions which has about a 50% chance of including a "good" solution. In contrast, shape complementarity would be more suitable to generate a list for subsequent refinement. The predicted complexes superimposed on the native structures are shown in Figure 8.8. The first "good" solution ranked on shape complementarity is shown.
To our knowledge, this work is the only systematic study of a simulation that performed a global search of docking proteins to DNA. The results show a somewhat poorer level of discrimination that we and other groups
Continue reading here: Lam
Was this article helpful?