The basic terms and concepts in the LabKey Server experiment framework are taken from the Functional Genomics Experiment (FuGE) project. The xar.xml format only encompasses a small subset of the FuGE object model, and is intended to be compatible with the FuGE standard as it emerges. More details on FuGE can be found at http://fuge.sourceforge.net
Objects that Describe an Experiment
The LabKey Server experiment framework uses the following primary objects to describe an experiment.
- Sample or Material: These terms are synonyms. A Sample object refers to some biological sample or processed derivative of a sample. Examples of Sample objects include blood, tissue, protein solutions, dyed protein solutions, and the content of wells on a plate. Samples have a finite amount and usually a finite life span, which often makes it important to track measurement amounts and storage conditions for these objects. Samples can be included in the description of an experiment as the input to a run. The derivation of Samples can be tracked.
- Sample Type A Sample Type is a group of Samples accompanied by a suite of properties that describe shared characteristics of all samples in the group.
- Data: A Data object refers to a measurement value or control value, or a set of such values. Data objects can be references to data stored in files or in database tables, or they can be complete in themselves. Data objects can be copied and reused a limitless number of times. Data objects are often generated by instruments or computers, which may make it important to keep track of machine models and software versions in the applications that create Data objects.
- Protocol or Assay: These terms are synonyms. A Protocol object is a description of how an experimental step is performed. A Protocol object describes an operation that takes as input some Sample and/or Data objects, and produces as output some Sample and/or Data objects. In LabKey Server, Protocols are nested one level--an experiment run is associated with a parent protocol. A parent protocol contains n child protocols which are action steps within the run. Each child protocol has an ActionSequence number, which is an increasing but otherwise arbitrary integer that identifies the step within the run. Child protocols also have one or more predecessors, such that the outputs of a predecessor are the inputs to the protocol. Specifying the predecessors separately from the sequence allows for protocol steps that branch in and out. Protocols also may have ParameterDeclarations, which are intended to be control settings that may need to be set and recorded when the protocol is run.
- ProtocolApplication: The ProtocolApplication object is the application of a protocol to some specific set of inputs, producing some outputs. A ProtocolApplication is like an instance of the protocol. A ProtocolApplication belongs to an ExperimentRun, whereas Protocol objects themselves are often shared across runs. When the same protocol is applied to multiple inputs in parallel, the experiment run will contain multiple ProtocolApplications object for that Protocol object. ProtocolApplications have associated Parameter values for the parameters declared by the Protocol.
- ExperimentRun: The ExperimentRun object is a unit of experimental work that starts with some set of input materials or data files, executes a defined sequence of ProtocolApplications, and produces some set of outputs. The ExperimentRun is the unit by which experimental results can be loaded, viewed in text or graphical form, deleted, and exported. The boundaries of an ExperimentRun are up to the user.
- RunGroup or Experiment: These terms are synonyms. LabKey Server's user interface calls these entities RunGroups while XAR.xml files call them Experiments. A RunGroup is a grouping of ExperimentRuns for the purpose of comparison or export. The relationship between ExperimentRuns and RunGroups is many-to-many. A RunGroup can have many ExperimentRuns and a single ExperimentRun can belong to many RunGroups.
- Xar file: A compressed, single-file package of experimental data and descriptions. A Xar file expands into a single root folder with any combination of subfolders containing experimental data and settings files. At the root of a Xar file is a xar.xml file that serves as a manifest for the contents of the Xar as well as a structured description of the experiment that produced the data.
Relationships Between xar.xml Objects
At the core of the data relationships between objects is the cycle of ProtocolApplications
and their inputs and outputs, which altogether constitute an ExperimentRun.
- The cycle starts with either Sample and/or Data inputs. Examples are a tissue sample or a raw data file output from an LCMS machine.
- The starting inputs are acted on by some ProtocolApplication, an instance of a specific Protocol that is a ProtocolAction step within the overall run. The inputs, parameters, and outputs of the ProtocolApplication are all specific to the instance. One ProtocolAction step may be associated with multiple ProtocolApplications within the run, corresponding to running the same experimental procedure on different inputs or applying different parameter values.
- The ProtocolApplication produces sample and/or data outputs. These outputs are usually inputs into the next ProtocolAction step in the ExperimentRun, so the cycle continues. Note that a Data or Sample object can be input to multiple ProtocolApplications, but a Data or Sample object can only be output by at most one ProtocolApplication.
The relationships between objects are intrinsically expressed in the relationships between tables in the LabKey Server database as shown in the following diagram:
Experiment Run Graph Notation
Experiment run graphs are presented using the following graphical notation representing the flow of the experiment.