Enhanced intensity-based clustering of isomorphous multi-crystal data sets in the presence of subtle variations

Thompson, A.J.; Beilsten-Edmands, J.; Tam, C.; Sanchez-Weatherby, J.; Sandy, J.; Mikolajek, H.; Axford, D.; Jaho, S.; Hough, M.A.; Winter, G.

doi:10.1107/S2059798325004589

Figure 4
(a) Unit-cell dimensions plotted as a histogram to demonstrate overlap between species. (b) Correlation-based HCA using the σ-weighted CC algorithm. (c) Cosine-angle HCA analysed in two dimensions using the σ-weighted CC algorithm. (d) Objective function residual (equation 1

) for each tested dimensionality. The automatically selected dimension is highlighted, which is the first dimension where the residual drops into the noise level as determined by the algorithm described in Section 2.2

. (e) Cosine-angle HCA analysed in three dimensions using the σ-weighted CC algorithm. (f) The optimized multi-dimensional cosym coordinates, projected in two dimensions, with the analysis performed in three dimensions. The coordinates have been rotated to align the axes with the two most significant eigenvectors derived from principal component analysis. Data sets are coloured according to clusters identified using the OPTICS algorithm [bovine (orange), porcine (pink) and human (blue) insulin]. Dendrogram links have colours that are randomly allocated and are not representative of groups. For the other 2D projections of the coordinate plot, see the supporting information.

STRUCTURAL
BIOLOGY

ISSN: 2059-7983

Volume 81| Part 6| June 2025| Pages 278-290

https://doi.org/10.1107/S2059798325004589

Open

access

Search IUCr Journals		doi		Advanced search
Author		volume	page