💡 PhosKinTime is an ODE-based modeling framework for analyzing phosphorylation dynamics over time. It integrates parameter estimation, sensitivity analysis, steady-state computation, and visualization tools to help researchers explore kinase-substrate interactions in a temporal context.
In cellular signaling pathways, a series of proteins are phosphorylated in an activation cascade that drives cellular responses. Understanding these post-translational modifications is critical.
Acknowledgements (Click to expand)
This project began from my master's thesis work in the Theoretical Biophysics group, now
the Klipp-Linding Lab, at Humboldt-Universität zu Berlin. The
submitted thesis focused on kinopt, while related components of the broader modelling framework were developed in
parallel during that period and continued afterwards.
The implemented version of kinopt was developed as the core thesis-stage component. The tfopt and protwise
components were developed by me during the same broader project period and subsequent continuation of the work. The
networkmodel component was later designed and implemented independently after that period as an extension of the
original modelling framework.
The initial distributive, successive, and combinatorial ODE formulations came from the thesis-stage modelling framework, while the saturation model was added independently at a later stage as an extension of the system.
The subpackage tfopt is an optimized and extended derivative
of original work by my
colleague Julius Normann, adapted with permission.
I am grateful to Ivo Maintz for generous technical support with server access, package experimentation, and computational setup.
PhosKinTime uses ordinary differential equations (ODEs) to model phosphorylation kinetics and supports multiple mechanistic hypotheses, including:
- Distributive Model: Phosphorylation events occur independently.
- Successive Model: Phosphorylation events occur sequentially.
- Combinatorial Model: Phosphorylation events occur in a hypercube lattice manner.
- Saturation Model: Phosphorylation dynamics follow saturating kinetics, where phosphorylation rates approach an upper limit as kinase input or substrate availability increases.
The repository includes executable educational Jupyter notebooks that demonstrate the core modeling workflows with tiny deterministic dummy data. They are designed for learning, CI execution, and quick smoke-testing of the current JAX/JAXopt/Diffrax implementation.
| Notebook | Module | What it demonstrates |
|---|---|---|
notebooks/01_kinopt_educational_workflow.ipynb |
kinopt |
Kinase/phosphosite-style input tables, KinOpt preprocessing arrays, alpha/beta constraints, local optimization, a ranked multistart solution ensemble, parameter export, and fit plots. |
notebooks/02_tfopt_educational_workflow.ipynb |
tfopt |
TF-target regulatory networks, TF protein/phosphosite effect matrices, constrained local optimization, ranked multistart outputs, regulatory-effect visualization, and saved tables. |
notebooks/03_protwise_educational_workflow.ipynb |
phoskintime.protwise |
Protein-wise ODE modeling with mRNA/protein/phosphosite modalities, mode-aware fitting logic, Diffrax-based ODE solving, JAXopt parameter estimation, multistart ranking, residual plots, and CSV/JSON exports. |
notebooks/04_networkmodel_educational_workflow.ipynb |
phoskintime.networkmodel |
Network-level multimodal data handling, adjacency construction, alpha/beta projection utilities, missing-modality cases, Diffrax/JAX-based solving, local constrained optimization, ranked multistart outputs, mode exports, and network/parameter plots. |
Run the notebooks interactively:
jupyter lab notebooks/This package is distributed under the BSD 3-Clause License.
See the LICENSE file for full details.