Design and construction of the PredikinDB database. (a) Illustration showing how a UniProt entry is parsed to link protein kinase sequences (names in bold) with phosphorylation sites. (b) PredikinDB table schema showing links between fields. Field headers in italics are primary keys. Abbreviations in parentheses indicate the UniProt line from which the field was derived. For clarity, 38 fields containing key protein kinase residues used in substrate prediction are summarised as one field.