Skip to main content
Fig. 6 | BMC Bioinformatics

Fig. 6

From: Prop3D: A flexible, Python-based platform for machine learning with protein structural properties and biophysical data

Fig. 6

The CATH-inspired hierarchical structure of Prop3D. The inherently hierarchical structure of CATH (A) is echoed in the design schema underlying the Prop3D-20sf dataset (B), as illustrated here. Prop3D can be accessed as an HDF5 file seeded with the CATH hierarchy for all available superfamilies. For clarity, an example of one such superfamily is the individual H-group 2.60.40.10 (Immunoglobulins), shown here as the orange sector (denoted by an asterisk near 4 o’clock). Each such superfamily is further split into (i) the domain groups, with datasets provided for each domain (atomic features, residue features, and edge features), as delineated in the upper-half of (B), and (ii) pre-calculated data splits, shown in the lower-half of (B), which exist as hard-links (denoted as dashed green lines) to domain groups. (The ‘sunburst’ style CATH diagram, from http://cathdb.info, is under the Creative Commons Attribution 4.0 International License.)

Back to article page