of PIP3 signalling Vladimir Kiselev LeNovere Group — Babraham Institute Signaling ISP Seminar, September 2013 Vladimir Kiselev Computational Biology of PIP3 signalling
pathway Phosphoinositides Systems Biology Properties Interactions both on the cell membrane (2D) and in the cytoplasm (3D) Kinase and phosphatase activities Phosphoinositide (PIP2, PIP3 etc.) complexity Vladimir Kiselev Computational Biology of PIP3 signalling
pathway Phosphoinositides Systems Biology Isoform Complexity Figure from www.biochemsoctrans.org Each PI species has 6 isoforms In total 8 ∗ 6 = 48 isoforms Vladimir Kiselev Computational Biology of PIP3 signalling
pathway Phosphoinositides Systems Biology Computational Modeling Has been widely used in different biological disciplines Has been proved to be robust and reliable Helps tackling the complexity problem Vladimir Kiselev Computational Biology of PIP3 signalling
Model Predictions Kinetic Modeling Deterministic – use a set of ODE to describe evolution of the system Assuming that the system is well stirred and spatially homogeneous Stochastic – more general approach directly taking into account system fluctuations Is required when number of particles in the systems is small Others (e.g. Rule Based Modeling) Vladimir Kiselev Computational Biology of PIP3 signalling
Model Predictions Mass-Spec Data PI, PIP, PIP2, PIP3 concentrations were measured using mass-spectrometry Impossible to distinguish between several species [PI(4, 5)P2] [PI(3, 4)P2] [PI(4, 5)P2] [PI(3, 5)P2] [PI(4)P] [PI(3)P] [PI(4)P] [PI(5)P] Vladimir Kiselev Computational Biology of PIP3 signalling
Model Predictions Analysis Figure 4: Area under PIP3 curve minus steady state level. Model is more sensitive to parameters which curve goes higher (or lower) towards left or right Vladimir Kiselev Computational Biology of PIP3 signalling
Model Predictions Parameter Estimation 1000 runs Best 10 runs were considered mutant, but also an artificial experiment which was used to set steady state levels of PIP3 to their values. Otherwise, steady state level of PIP3 would be much higher or lower than zero-time experimental points in WT and PTEN experiments. Set of parameters that gives the best objective value, the best fit for experi- mental values, is given in Table 1, together with sensitivity analysis results; the fit is shown in Fig. 6. Summary of the best three sets of parameters (the one from Table 1, and two other sets which give very similiar objective values) is given in Table 2. The first two sets are somewhat similiar, but the third one di↵ers greatly in two parameters, which hit their upper boundary. Only the best set of parameters is kept and used later. Parameter Ranking value From To Source Value pi3k 1+ - - literature 0.2 pip3 pi45p2 production.k1 1 0 1100 estimated 283.069 pi3k activation.Kact 0.4092065232 0 1000 estimated 468.605 pi3k activation.k 0.3528758625 0 10000 estimated 4510.44 rec 0.3332101512 - - literature 0.15 pip3 degradation.k1 0.3161319637 0 12 estimated 0.0638764 egfr internalization.k1 0.2827946551 - - literature 0.0055 pi3k deactivation.k1 0.2721791102 1 1000 estimated 203.269 pi3k pi45p2 binding.k1 0.2721195867 0 100 estimated 8.44855 pten 0.2454299223 0 1 estimated 0.502768 pip3 pten binding.k1 0.2454217573 0 10 estimated 0.103891 egfr production.v 0.2136297575 0 0.0003 estimated 0.000181083 pi3k pi45p2 binding.k2 0.1962720248 0 3000 estimated 632.48 pi3k activation basal.k1 0.1763170064 - - x 0.108887 egf binding.k2 0.1549547047 - - literature 0.022 egf binding.k1 0.1502826433 - - literature 29 egfr dimerization.k2 0.1486302434 - - literature 0.3 egfr phosphorylation.k2 0.1470041127 - - literature 0.08 egfr dimerization.k1 0.1465098154 - - literature 10 egfr phosphorylation.k1 0.1425294984 - - literature 1.33 pi45p2 pip3 production.k1 0.1240844428 - - x 7.14235 pip3 pten binding.k2 0.0369564535 - - x 0.00941346 pi45p2 - - - experimental 115 lig - - - experimental 0.00157 Table 1: Parameters: ranking, intervals, final values 6 Vladimir Kiselev Computational Biology of PIP3 signalling
Model Predictions Results Figure 6: Results of parameter estimation: output of the model with the best set of parameters compared to experimental data. Vladimir Kiselev Computational Biology of PIP3 signalling
Model Predictions FRET Data PI(3,4)P2 concentration was identified at the University of Dundee Vladimir Kiselev Computational Biology of PIP3 signalling
Model Predictions Future Plan Adding spatial dimension Stochastic reactions Single molecule representations rather than populations Vladimir Kiselev Computational Biology of PIP3 signalling
Data Preliminary Analysis Time-Course Analysis Brief Intro Figure from www.discovery.lifemapsc.com Vladimir Kiselev Computational Biology of PIP3 signalling
Data Preliminary Analysis Time-Course Analysis Read Alignment Figure from www.arrayserver.com Vladimir Kiselev Computational Biology of PIP3 signalling
Data Preliminary Analysis Time-Course Analysis DE Analysis Tool DESeq package (from Bioconductor) – takes into account all replicates FDR of 1% Vladimir Kiselev Computational Biology of PIP3 signalling
Data Preliminary Analysis Time-Course Analysis Future Plan Identify unique EGF stimulation related genes Vladimir Kiselev Computational Biology of PIP3 signalling
Mouhannad Malek, Veronique Juvin Len Stephens, Phill Hawkins Nicolas Le Novere and the group Nicholas Luscombe and the LRI group Marija Jankovic, Sven Bergmann, Anne Segonds-Pichon, Simon Andrews Money BBSRC grant Babraham Institute Note: this presentation was made with L ATEX, Beamer + Wiki2Beamer Source code: https://github.com/wikiselev/department-seminar Vladimir Kiselev Computational Biology of PIP3 signalling