wiki:Concepts

Simulation Concepts

This is a combination of logical concepts and limitations that determine how Simulation works. These rules are enforced through the application and, to some extent, the database.

The main objects in the system are experiments, simulations, spectra, pulse sequences and metabolites. Experiments are the primary objects; everything else is secondary. Here's how they're related --

  • Each experiment has zero to many simulations. Simulations are the whole point of an experiment, and there's not much to an experiment besides the metadata that defines the simulations. Since entering the experiment metadata is pretty trivial, we don't let users save experiments that define zero simulations. Experiments with zero simulations can exist, but only in memory. They are never saved to the database or an export file.
  • Each experiment makes use of and refers to exactly one pulse sequence, but the experiment may define one or more timing sets for the pulse sequence.
  • Each simulation creates one spectrum.
  • Each spectrum has zero or more lines. Zero is an unusual case, but possible.
  • Each spectrum line has a one PPM, area and phase value in it.

We expect users to share data via Simulation's export and import functions. For this reason, several of Simulation's objects (experiments, pulse sequences and metabolites) have universally unique ids (UUIDs) rather than just ordinary integer ids.

The "Big Three"

Experiments, metabolites and pulse sequences have a number of things in common.

  • Simulation assigns each object a UUID when it's created and the UUID remains the same for the life of the object.
  • Simulation also sets the created timestamp when the object is created. That doesn't change either.
  • Their names must be unique among their kin. In other words, metabolite names must be unique among metabolites, experiment names must be unique among experiments, etc. It's OK (although perhaps a little strange), if for example a metabolite has the same name as a pulse sequence.
  • As of this writing, valid characters in a name are dash, dot (period), slash, space and alphanumeric characters plus underscore (as enumerated by the regex \w; see Python re module docs). This rule is goverened by the regular expression NAME_REGEX in constants.py.
  • Metabolites and pulse sequences used in (i.e. referred by) an existing experiment may not be deleted. See States (below).
  • All of these objects may be cloned. Cloning an object makes a copy of it with a brand new UUID. It's as if you clicked the "new" button and typed in fresh data yourself. If you've read the section on States (below), you'll understand when I say that since the result of a cloning is new, it's also private and not in use (and therefore not frozen).

Experiments

Experiments are the main focus of the Simulation application. An experiment's raison d'etre is to run a set of simulations. This set of simulations is the experiment's results space.

Currently, that space is defined by two to four nested loops. The first loop covers the list of metabolites the user has involved in the experiment. The other one, two or three loops are user-defined lists of numbers.

The attachment results_space.png is a visual representation of a 3D results space (one set of metabolites and two lists of user-defined numbers).

Simulations themselves know nothing about one another and are agnostic to the order in which they're run. Thus, while the existing code is geared towards generating a very regular results space that we iterate over in a very straightforward order, more complex result spaces and iteration orders are possible. The sky's the limit, really, provided you can dream up a GUI that allows users to describe the results space.

There's more about the results space in the section on simulations.

  • Once an experiment has been saved, the following attributes become read-only: pulse sequence, investigator, user parameters, b0, isotope, peak_search_ppm_low, peak_search_ppm_high, blend_tolerance_ppm, blend_tolerance_phase.
  • One can associate additional metabolites with an experiment, but once it is associated and the experiment is saved, the metabolite remains with the experiment forever. In other words, a metabolite can't be removed from a saved experiment.

  • An experiment's b0 value is always stored in megahertz.

Pulse Sequences

  • A loop without a label name will be treated as a "non-loop". That is, the loop control will not show up in the Visualization widget. All loops used in a pulse sequence must have a label.
  • A user defined parameter must have a name and default value listed. The default values should be such that the pulse sequence will run with only default values being used, but Simulation will not check for this, only suggest it be so in the documentation.
  • The pulse sequence sequence_code field cannot be empty. Simulation is not responsible for checking to see that the code is syntactically correct, working PyGamma code (although that would be nice).
  • The pulse sequence binning_code field is populated with default binning code. The user can replace this with their own code if they wish to. Simulation is not responsible for checking to see that the code is syntactically correct, working PyGamma code (although that would be nice).

Metabolites

The main fields of a metabolite are UUID and name. They also have one or more spins and zero or more J couplings. The number of J couplings will depend on the number of spins, but may contain all zeros depending on the underlying physics.

Spins and J Couplings

Spins are always associated with a specific metabolite. Their main attributes are isotope (e.g. 1H) and chemical shift (a float).

A J Coupling describes the relationship between two spins. It has a float value (in Hertz) and the two spins to which it refers are always on the same metabolite.

  • Each metabolite has one or more spins.

  • The order of the spins isn't important. For sanity's sake, we always retain the order that the user provided.

  • Our app imposes a somewhat arbitrary limit of 14 spins per metabolite. As of this writing, the limit derives from only one place -- a constant in the metabolite editing dialog. In theory, one can change that constant and Simulation will take it in stride.

  • Each metabolite has zero or more J couplings. There's one J coupling per unique combination of spins, which works out to be the sum(n) where n == the number of spins. All couplings can be set to zero or any other float number.

For instance, if there's three spins, there are three J couplings (1-2, 1-3, and 2-3). If there's, four spins, there's six J couplings (1-2, 1-3, 1-4, 2-3, 2-4 and 3-4). The following very long line of Python will list the spin count-J coupling relationship.

print "spins\tj-coups\n" + "\n".join(["%d\t%d" % (i, sum(range(i))) for i in range(1, 15)])

That prints something very much like the table below.

spinsJ couplings
10
21
33
46
510
615
721
828
936
1045
1155
1266
1378
1491

Isotopes

The isotopes associated with each spin are restricted to those on the following list (which is reflected in the isotopes table in our database):

 1H   
 2H   
 3H   
 6LI  
 7LI  
10B  
11B  
13C  
14N  
15N  
17O  
19F  
29SI 
31P  

Note that the list is not ordered alphabetically but by atomic mass.

Simulations

In the context of this application, the word "simulation" has a specific meaning that clashes a little with the real world.

Recall that an experiment runs a loop over two to four dimensions and that's called the results space. In the real world, people are apt to call the entire results space a simulation but we're calling it an experiment. In the context of this app, we use the term slightly differently: each point in the results space is one simulation.

  • Each simulation has a value for metabolite, dimension 1, dimension 2, and dimension 3.

  • Each combination of these values is unique within an experiment.

  • Each simulation produces a spectrum as output.

See the attachment results_space.png for a visual representation of the simulations run by an experiment.

Spectra

A spectrum is the output from a simulation. There's exactly one spectrum produced by each simulation. Since an experiment comprises a 3D block of simulations, then the output of an experiment is a 3D block of spectra.

  • A spectrum is composed of a number of lines.

  • Each line has the values PPM, area and phase.
  • When visualized graphically, the PPM values represent the X axis and so we always sort lines by PPM.

States Within Simulation

Active and Deactivated (Metabolites)

Metabolites are active by default. Deactivating a metabolite reduces clutter by taking it "out of service". Inactive metabolites don't show up in the list of available metabolites when creating a new experiment. That's the only effect of active/inactive status.

You can change a metabolite from active to inactive and vice versa any time you like.

Private and Public (Experiments, Metabolites and Pulse Sequences)

All of these objects start life private. That means they're only in your database; no one else has seen them.

Once exported, objects become public. That means that their definition has been shared with the world. Public objects are frozen (see below).

Once a private object has become public, it can never become private again. Cloning, however, will create a new, private object with exactly the same properties (but a different UUID).

In Use (Metabolites and Pulse Sequences)

When you select a metabolite or pulse sequence for use in an experiment, that experiment refers to the object for as long as the experiment exists in your database. Metabolites and pulse sequences that are referred to by an experiment are in use by that experiment.

Objects that are in use may not be deleted and are frozen (see below).

There's no limit on the number of experiments that can refer to a given metabolites or pulse sequence.

Once all of the experiments referring to an object are deleted, the object is no longer in use.

Frozen (Experiments, Metabolites and Pulse Sequences)

Frozen objects are mostly uneditable -- only the name and comments can be changed. Objects are frozen for one of two reasons.

In use objects (metabolites and pulse sequences) are frozen because they're referred to by an experiment, and changing the underlying objects that the experiment uses would corrupt the experiment.

Public objects are frozen because once you've shared an object with others (or you've imported an object that they've shared with you), you need to be able to trust that you're talking about exactly the same object.

Here's a table summarizing when an object is frozen.

Private?In Use?Frozen?
YesNoNo
YesYesYes
No (public)NoYes
No (public)YesYes

The in use concept doesn't apply to experiments, so for them only private/public matters but the connection to frozen is still the same.

Note that frozen only refers to whether or not the fundamental attributes of the object can be edited. It doesn't affect whether or not it can be deleted.

Last modified 9 years ago Last modified on Jul 1, 2010, 4:06:28 PM

Attachments (2)

Download all attachments as: .zip