Protein kinases control cellular decision processes by phosphorylating specific substrates. Proteome-wide mapping has identified thousands of in vivo phosphorylation sites. However, systematically resolving which kinase targets each site is presently infeasible, due to the limited specificity of consensus motifs and the potential influence of contextual factors, such as protein scaffolds, localisation and expression, on cellular substrate specificity. We have therefore developed a computational method, NetworKIN, that augments motifs with context for kinases and phosphoproteins. This can pinpoint individual kinases responsible for specific in vivo phosphorylation events and yields a 2.5-fold improvement in the accuracy with which phosphorylation networks can be constructed. We show that context provides 60–80% of the computational capability to assign in vivo substrate specificity. Applying this approach to a DNA damage signalling network, we extend its cell-cycle regulation by showing that 53BP1 is a CDK1 substrate, show that Rad50 is phosphorylated by ATM kinase under genotoxic stress, and suggest novel roles of ATM in apoptosis. Finally, we present a scalable strategy to validate our predictions and use it to support the prediction that BCLAF1 is a GSK3 substrate.