Fig. 3

The flowchart of obtaining candidate clusterings. The figure depicts the procedure of obtaining the set of candidate clusterings. Based on the kernel matrix (calculated in the previous step), for each value of m, the protein structure is first split into m clusters (using either kernel k-means or spectral clustering). Next, the divided alpha-helices with a maximum size of MHS are merged with their adjacent segments. In the subsequent steps the short and excessive segments (with respect to the MSS and SDR, respectively) are removed. The process is then repeated with increasing m until the generated partitioning is rejected according to the MDS parameter (designated as the accept step in the flowchart)