How ProcessDB is organized¶

Everything in ProcessDB is built from a small set of objects, and every panel, menu, and dialog is named after them. This page defines them once, so the rest of the application reads as itself.

The building blocks of a mechanism¶

Molecule. A chemical species — ATP, glucose, a named protein. A molecule has no location and no concentration; it is just the identity of a substance. Molecules are shared across every model in the database, so "ATP" means the same thing everywhere.

Complex. A molecule assembled from other molecules, with a count of each part (for example a heterodimer, or a protein with two bound ligands). Complexes behave like molecules everywhere else in ProcessDB.

Place. A compartment — cytosol, mitochondrial matrix, extracellular space, a plasma pool. Places are also shared across models.

State. A molecule in a place. "Glucose in Cytosol" and "Glucose in Blood" are two different states of one molecule. A state is the variable that gets a differential equation, so this is the level at which concentrations, initial conditions, and measurements live.

Process. A step that changes states: a reaction, a transport event, a degradation, a synthesis. Each process has participants:

• Reactants are consumed, each with a stoichiometric coefficient.
• Products are produced, likewise with coefficients.
• Modifiers change the rate without being consumed: enzymes, catalysts, transporters, channels, mediators, activators, inhibitors, and general regulators. A role says that a state regulates a process; how an activator or inhibitor enters the equation — competitively, non-competitively, as an essential or a non-essential activator — is a property of the rate law, chosen when the law is built. See How regulators enter a modular law.

A process with no reactants is a synthesis; one with no products is a degradation. See Processes and roles for how the roles are drawn and edited.

The four layers¶

Model — the hypothesis¶

A model is a set of processes, together with the states and places they involve, a diagram layout, optional subdiagrams, and notes. It is purely structural: it says what can happen, not how fast. Two people looking at the same model are looking at the same claim about the biology.

Because a model carries no numbers, it can be shared, re-used, and argued about on its own terms — and the same model can support several competing quantitative treatments.

Model Realization — the kinetics¶

A Model Realization (MR) makes a model quantitative. It adds, for the model as a whole:

• an active rate law for every process — an algebraic formula plus its parameters;
• an initial condition for every state;
• for states you mark as boundary states, a boundary formula giving the value at all times rather than a differential equation;
• optional MR functions: named algebraic definitions available to every simulation built on this realization.

One model can have many realizations. That is the mechanism by which you compare mass-action against Michaelis–Menten kinetics, or two different inhibition schemes, without duplicating the diagram.

Experiment — the bench¶

An experiment records what was done and what was seen:

• State protocols and process protocols — the perturbations (a bolus addition, an infusion, a clamped concentration, a time course imposed on a flux, a discrete event);
• tagged molecules and their tracer protocols — labelled species and how the label was introduced or destroyed;
• measurements and data sets — the observed time courses, each measurement associated with a model variable, each data point carrying a weight.

Experiments are independent of any one model. The same experiment can be used to test several competing models, which is exactly what you want when comparing hypotheses.

Model of Experiment — the simulation¶

A Model of Experiment (MOE) is one Model Realization simulated under a chosen set of experiments. It is the object that actually runs. A MOE owns:

• the generated equation listing — the complete mathematical model, written out;
• parameter sets — named collections of parameter values you can switch between;
• MOE functions — algebraic definitions specific to this MOE and experiment;
• traced states and processes — which parts of the network carry each label;
• the start and end time of the simulation, and the graph layout you last used.

When a MOE includes several experiments, ProcessDB generates one copy of the state equations per experiment, so a single simulation covers every condition at once and a single parameter set has to explain all of them.

How a change propagates¶

Model  ──►  Model Realization  ──►  Model of Experiment  ──►  simulation
(structure)      (kinetics)            (+ experiments)          (results)

Edits flow downhill. Adding a process to the model, changing a rate law, or adding an experiment to a MOE all make that MOE's equations out of date; ProcessDB says so with a banner above the equation listing and offers to regenerate them. Nothing is regenerated behind your back while you are working, so what you see is always the listing that produced the results you are looking at.

Identifiers you will see¶

Generated equations refer to states and processes by symbol rather than by name, because names change and are not always unique:

Wherever a listing or table shows these symbols there is a Display Names toggle that swaps them for the state and process names, and you can double-click a symbol to jump to that object in the diagram.