Accurate prediction of kinase-substrate networks using knowledge graphs

Abstract

Phosphorylation of specific substrates by protein kinases is a key control mechanism for vital cell-fate decisions and other cellular processes. However, discovering specific kinase-substrate relationships is time-consuming and often rather serendipitous. Computational predictions alleviate these challenges, but the current approaches suffer from limitations like restricted kinome coverage and inaccuracy. They also typically utilise only local features without reflecting broader interaction context. To address these limitations, we have developed an alternative predictive model. It uses statistical relational learning on top of phosphorylation networks interpreted as knowledge graphs, a simple yet robust model for representing networked knowledge. Compared to a representative selection of six existing systems, our model has the highest kinome coverage and produces biologically valid high-confidence predictions not possible with the other tools. Specifically, we have experimentally validated predictions of previously unknown phosphorylations by the LATS1, AKT1, PKA and MST2 kinases in human. Thus, our tool is useful for focusing phosphoproteomic experiments, and facilitates the discovery of new phosphorylation reactions. Our model can be accessed publicly via an easy-to-use web interface (LinkPhinder).

Accurate, high-coverage assignment of in vivo protein kinases to phosphosites from in vitro phosphoproteomic specificity data

“…f) IV-KAPhE outperforms the simpler PSSM-based and Naïve Bayes+ methods as well as other previously published methods in kinase-substrate assignment of an external validation set. Points indicate the scores for simple assignments (GPS) or the scores at nominal cutoffs for quantitative predictions (cutoffs—IV-KAPhE: 0.5, PSSM: 0.75, Naïve Bayes+: 0.5, LinkPhinder: 0.672 [ 27 ], NetworKIN 3.0: 1.0 [ 40 ]). Error bars show the 95% confidence intervals at these points.…”

“…I furthermore evaluated the assignments for these kinase-site pairs made by phosphoproteome-backed PSSMs, Naïve Bayes+, and three other, previously published tools with similar kinomic scope or model architecture: NetworKIN 3.0 [ 40 ], GPS 5.0 [ 17 ], and LinkPhinder [ 27 ]. NetworKIN and GPS were run in-house with their default settings, whereas the LinkPhinder scores produced by the authors were used.…”

“…NetworKIN and GPS were run in-house with their default settings, whereas the LinkPhinder scores produced by the authors were used. For further evaluation, the published stringent LinkPhinder cutoff of 0.672 [ 27 ] and the nominal NetworKIN cutoff of 1.0 [ 40 ] were used.…”

“…The problem has instead changed to one of kinase-phosphosite assignment . Accordingly, new methods have emerged, based on models such as support vector machines [ 21 , 22 ], multiple kernel learning [ 23 ], Naïve Bayes [ 24 ], networks [ 25 , 26 ], and knowledge graphs [ 27 ]. Classical scoring matrices are often still used within these or related methods to model kinase substrate specificity [ 23 , 25 , 26 , 28 ].…”

Systematic identification of ALK substrates by integrated phosphoproteome and interactome analysis

“…In these analyses, only a small fraction (often less than 5%) of the quantified phosphorylation sites were used, mainly because much less information is available on known kinase–substrate relationships (KSRs) than on phosphorylation sites quantified by MS ( Needham et al, 2019 ). Various strategies have been used to obtain high-quality KSR data on a large scale; these approaches include informatics approaches ( Invergo et al, 2020 ; Nováček et al, 2020 ), approaches using in vitro kinase assays ( Knebel et al, 2001 ; Newman et al, 2013 ; Sugiyama et al, 2019 ), chemical proteomics approaches using kinase inhibitors in vivo ( Hijazi et al, 2020 ; Watson et al, 2020 ), and biochemical and genetic approaches combining proximity-dependent biotinylation (BioID)-based interactome and phosphoproteome analyses ( Cutler et al, 2020 ; Niinae et al, 2021 ). The informatics approach relies on public phosphoproteome data or KSR data as input data; thus, predicting new substrate candidates for a kinase with limited KSR data is difficult by this approach.…”