Skip to content

Experiments

An experiment in ProcessDB is the record of one bench experiment: the perturbations that were applied, any labelled species that were introduced, and the time courses that were measured.

Experiments are stored independently of any model. That is deliberate: the same experiment can be used to test several competing models, and a model can be tested against a whole series of experiments at once. An experiment is evidence; a model is a claim; the Model of Experiment is where they meet.

Creating an experiment

Press New Experiment in the Experiments browser. Open it and give it a name, a description, and a date; Rename and Change Date are also on the right-click menu, in the browser and on the tab.

The experiment tab

A summary sits across the top — name, description, and date — beside a Target Models list showing the models this experiment is currently being used to test. That list fills in automatically from the MOEs that include the experiment; you can also add a model to it by hand, to keep a model you are about to test in view while you enter the protocols.

The rest of the tab is four sub-tabs. A green dot on a tab means it has content, so you can see at a glance what has been entered:

Tab Holds
State Protocols What was done to individual species — additions, infusions, clamps, imposed time courses, discrete events
Process Protocols What was done to a flux — imposed rates or measured flux time courses
Tracer Protocols Labelled molecules and how the label was introduced or destroyed
Experimental Measurements and Data What was measured, and the numbers

See Protocols, Tracers and tagged molecules, and Measurements and data.

How an experiment reaches a simulation

Add the experiment to a Model of Experiment — press + on the MOE's Experiments in MOE list, or drag the experiment there from the browser. When the MOE's equations are generated, ProcessDB writes:

  • one copy of the state equations per experiment, with each symbol tagged by the experiment ID (S123_7 is state 123 in experiment 7);
  • the protocol equations for that experiment, layered onto those copies;
  • the tracer equations for any tagged molecules the experiment uses.

Everything runs in a single simulation. That means one parameter set has to explain every experiment in the MOE at once, which is exactly the constraint that makes a fit meaningful. It also means adding an experiment to a MOE changes its mathematics — ProcessDB will tell you the equations are out of date and offer to regenerate them.

A note on scope

An experiment describes conditions and observations, not kinetics. Nothing entered here changes a rate law. When an experiment seems to call for a different rate constant, that is what a second Model Realization is for — or a protocol that imposes the altered quantity directly.