Figure 1. A flowchart of meta-analysis loop.

Distance matrices are calculated from N related PDB entries in form of Cartesian coordinates. The lower triangles of the distance matrices are assembled into columns of matrix A. The first a few significant left singular vectors in U resulting from SVD are reconstructed to form decomposed lower triangles. The corresponding singular values and right singular vectors are presented in a multidimensional conformational space. Any given location in this space leads to a recomposed lower triangle by a linear combination of the significant lower triangles. Finally, a hypothetical structure is determined by distance geometry using all constraints in the recomposed lower triangle.