|
Fig 1. Structural and evolutionary analysis of X. laevis phosphosites. A) A total of 2636 non-redundant phosphorylation sites were compiled from the sites determined here and those collected from a previous study [29]. We determined the conservation of these 2636 phosphorylation sites across the 13 other species and obtained structural models for 518 of these sites. B) The all-atom residue relative surface accessibility was compared for all phospho-acceptor residues, phosphosites not conserved or conserved in at least one other species with available phosphorylation data. C) The fraction of X. laevis sites with a known function in human increases with the degree of conservation. X. laevis sites not conserved in human were excluded from this analysis. D) Example comparative models with highly conserved phosphorylation sites. The phosphorylation site is highlighted in red. For the NDP kinase A, the structure represents the homo-oligomeric complex. One of the subunits is indicated in blue, with the phosphosite position in red and the substrate in the ball-and-stick representation.
doi:10.1371/journal.pcbi.1004362.g001
|
|
Fig 2. Phosphosites in solvent inaccessible positions may be predictive of conformational flexibility.
A) A small fraction of phosphosites (approximately 20%) was observed to be at solvent inaccessible positions (defined here as <20% all-atom RSA). The distribution of phosphosite RSA and the fraction of low accessibility sites do not vary significantly as a function of target-templates sequence identity nor phosphosite conservation. B) For phospho-acceptor residues (Serine, Threonine and Tyrosine) modeled independently based on more than one template structure, we compared the RSA values obtained from different models. These values are highly correlated, although some sites showed large variability in predicted accessibility, potentially indicating regions of conformational flexibility. C) We compared the changes in RSA in different models for phospho-acceptor residues, phosphoryation sites not known to be conserved, those conserved in at least 1 other species, phosphosites predicted by DISOPRED to be in ordered or disordered regions. D) Examples of phosphosites found in positions that show a large change in accessibility in two templates and are poorly accessibility in one of the templates. The phosphosite position is highlighted in red and the models in orange have a higher RSA for the phosphosite position when compared to the model in cyan.
doi:10.1371/journal.pcbi.1004362.g002
|
|
Fig 3. Conservation of putative cell-cycle related kinase interactions is predictive of known and/or cell-cycle related kinase-target interactions.
A) The number of predicted kinase target sites and proteins associated with cell-cycle kinases selected for analysis in X. laevis. We tested if the degree of conservation of kinase-interactions was predictive of known interactions; enriched in proteins that are phospho-regulated in the cell cycle; and genes known to cause cell cycle phenotypes when knocked down. B) ROC curves measuring the accuracy for kinase-interaction predictions. C) Enrichment over random prediction for the 3 tested features. D) Predicted kinase interactions conserved in 7 or more species are shown, highlighting known interactions, proteins phospho-regulated during the cell cycle and genes causing cell cycle phonotypes. The edge thickness is proportional to the degree of conseravtion for the predicted kinase-protein interactions. A list of these interactions is provided in S4 Table.
doi:10.1371/journal.pcbi.1004362.g003
|
|
S1 Fig. Enrichment of functional phosphorylation events as a function of conservation.
For each X. laevis phosphosite we counted the number of species in which the orthologous peptide region is also phosphorylated. We excluded all phosphosites that are not also phosphorylated in human. We then calculated the fraction of sites that are known to play a functional role in human. The degree of conservation is found to enrich significantly for sites with a known function for all X. laevis sites as well as sites that are in ordered or disordered regions.
|
|
S2 Fig. Benchmark for kinase specificity position specific scoring matrices.
For 16 cell-cycle kinases with many known target sites (Akt, Atr, AurA, AurB, Cdk1, Cdk2, Cdk3, Cdk7, Chk1, Nek2, Nek6, Nek9, Plk1, Plk2, Plk3, Ttk), position specific scoring matrices (PSSMs) were derived for each kinase and benchmarked on a set of known target sites using a cross-fold validation 16. We show here the AROC curves for the cross validation for each kinase.
doi:10.1371/journal.pcbi.1004362.s002
|
|
S3 Fig. Median values for cross-validation AROC values for 16 cell-cycle kinases with many known target sites (Akt, Atr, AurA, AurB, Cdk1, Cdk2, Cdk3, Cdk7, Chk1, Nek2, Nek6, Nek9, Plk1, Plk2, Plk3, Ttk).
|
|
For 16 cell-cycle kinases with many known target sites (Akt, Atr, AurA, AurB, Cdk1, Cdk2, Cdk3, Cdk7, Chk1, Nek2, Nek6, Nek9, Plk1, Plk2, Plk3, Ttk), position specific scoring matrices (PSSMs) were derived for each kinase and benchmarked on a set of known target sites using a cross-fold validation 16. We show here the AROC curves for the cross validation for each kinase.
|
|
|
|
For 16 cell-cycle kinases with many known target sites (Akt, Atr, AurA, AurB, Cdk1, Cdk2, Cdk3, Cdk7, Chk1, Nek2, Nek6, Nek9, Plk1, Plk2, Plk3, Ttk), position specific scoring matrices (PSSMs) were derived for each kinase and benchmarked on a set of known target sites using a cross-fold validation 16. We show here the AROC curves for the cross validation for each kinase.
|
|
For 16 cell-cycle kinases with many known target sites (Akt, Atr, AurA, AurB, Cdk1, Cdk2, Cdk3, Cdk7, Chk1, Nek2, Nek6, Nek9, Plk1, Plk2, Plk3, Ttk), position specific scoring matrices (PSSMs) were derived for each kinase and benchmarked on a set of known target sites using a cross-fold validation 16. We show here the AROC curves for the cross validation for each kinase.
|
|
For 16 cell-cycle kinases with many known target sites (Akt, Atr, AurA, AurB, Cdk1, Cdk2, Cdk3, Cdk7, Chk1, Nek2, Nek6, Nek9, Plk1, Plk2, Plk3, Ttk), position specific scoring matrices (PSSMs) were derived for each kinase and benchmarked on a set of known target sites using a cross-fold validation 16. We show here the AROC curves for the cross validation for each kinase.
|
|
For 16 cell-cycle kinases with many known target sites (Akt, Atr, AurA, AurB, Cdk1, Cdk2, Cdk3, Cdk7, Chk1, Nek2, Nek6, Nek9, Plk1, Plk2, Plk3, Ttk), position specific scoring matrices (PSSMs) were derived for each kinase and benchmarked on a set of known target sites using a cross-fold validation 16. We show here the AROC curves for the cross validation for each kinase.
|
|
For 16 cell-cycle kinases with many known target sites (Akt, Atr, AurA, AurB, Cdk1, Cdk2, Cdk3, Cdk7, Chk1, Nek2, Nek6, Nek9, Plk1, Plk2, Plk3, Ttk), position specific scoring matrices (PSSMs) were derived for each kinase and benchmarked on a set of known target sites using a cross-fold validation 16. We show here the AROC curves for the cross validation for each kinase.
|
|
For 16 cell-cycle kinases with many known target sites (Akt, Atr, AurA, AurB, Cdk1, Cdk2, Cdk3, Cdk7, Chk1, Nek2, Nek6, Nek9, Plk1, Plk2, Plk3, Ttk), position specific scoring matrices (PSSMs) were derived for each kinase and benchmarked on a set of known target sites using a cross-fold validation 16. We show here the AROC curves for the cross validation for each kinase.
|