3.2 Global comparison in working dataset as compared to control
Next, the 895 cases from the bound versus unbound pairs of protein chains are subject to global comparison by computing the network dissimilarity score (NDS) and RMSD for structure deviation between backbone conformations. The information obtained from the working dataset is illustrated as a scatter plot as shown in Figure 1A. Similar information obtained in the control where only multiple structures of the same protein that are known to be single chain single domain is also plotted in the background of the scatter. The statistics of the data points corresponding to each of the datasets is marked with horizontal and vertical lines to depict the mean of the data points. The structure deviation in the control was about 0.34 Å with a standard deviation of about 0.3 Å, whereas the mean deviation of the working dataset is about 1.14 Å with a standard deviation of about 0.94 Å. By performing a KS test, it is also understood that the RMSD scores obtained in the working dataset is significantly much higher than what was observed in the control (p-value < 0.0001). Hence, the spread of the topological variation of the backbone is much larger when proteins undergo transient associations than the structural variability obtained from multiple conformers of the same protein.
Likewise, the network dissimilarity observed in the control had a mean of 0.112 ± 0.047 whereas that observed from the working dataset is 0.173 ± 0.065. Also, KS-test results (p-value < 0.0001) show that the dissimilarity in the networks is significantly higher in transiently binding proteins as compared to the dissimilarity in PSNs of multiple conformers of the same protein (control dataset). It is interesting to analyse those cases where the variability in the network arises without much change in the backbone conformation. The contribution to network dissimilarity specifically arises from variability in local sidechain conformations. The network comparison scores, and structure deviation of these cases are greater and lower than mean of the dataset, respectively and can be found clustered on the top left of the scatter shown in Figure 1A. 194 cases from the working dataset show such a trend (with RMSD < 1.14 Å and NDS > 0.173). A list of top twelve cases with such network alterations are listed in Figure 1B, a few highlighted cases are studied in detail.