Figure 1
From: Predicting functional sites with an automated algorithm suitable for heterogeneous datasets
Partitioning Around Medoids Clustering (PAMC) is used to partition PSZs between -1.0 and χ = -2.0 into two groups (signal and noise clusters). The results from the (A) raw and (B) sharpened PSZ datasets are shown. Red indicated sharpened data points. We demonstrate two common scenarios (left) triosephosphate isomerase and (right) arginyl-tRNA synthetase. In all cases, it is clear that sharpening the PSZ dataset allows one to more easily discern the number of distinctive potential signals under the partition boundary (gap) in question. (C) The effect of different sharpening ranges is demonstrated. PSZ ranges tested include -1.0 to -2.0 (red), -2.5 (dark orange), and -3.0 (light orange). In the case of triosephosphate isomerase the first two ranges give identical results. The ideal threshold is found to be -1.65 and -2.20 for triosephosphate isomerase and arginyl-tRNA synthetase, respectively.