Investigation

This document is intended to be a general overview of the OBI model of an investigation. It does not match the current version of OBI on all points -- there are some new classes and modifications of existing classes. The development notes below detail those changes.

Feedback would be appreciated. Please add comments or suggestions to the discussion page.

= General Model of a Biomedical Investigation =

At the heart of the Ontology for Biomedical Investigations is our model of an investigation. An investigation is a process with three parts:


 * 1) the planning stage, in which a study design is created.
 * 2) the study design execution stage, in which the steps of the study design are carried out.
 * 3) the #|investigation reporting stage, in which an #|investigation report is created.

The output of an investigation is the #|investigation report, which includes a statement of the conclusions of the investigation.

Most biological investigations involve collecting specimens, preparing them, generating data by taking measurements, and then working with data. Most medical investigations involve enrolling subjects, making an intervention, generating data by taking measurements, and then working with data. While we call people &quot;subjects&quot; and cells &quot;specimens&quot;, the processes can be modelled in very similar ways. Here area four types of process common to many study designs:


 * Biological Investigation
 * specimen creation (or collection)
 * material processing
 * assay
 * #|data processing
 * Medical Investigation
 * subject enrollment
 * #|subject processing
 * assay
 * #|data processing

Each of these steps in the study design has inputs and outputs. In the simplest case these processes form a chain where the output of one step is the input to the next step.



Specimen Creation (Collection)

 * input: unspecified (outside the scope of the OBI model of investigations)
 * output: specimen -- such as a protein, cell, organism, or population
 * examples:
 * collecting specimen from organism such as
 * taking a sputum sample from a cancer patient
 * taking the spleen from a killed mouse
 * collecting a urine sample from a patient
 * environmental material collection such as
 * taking 1 liter of surface ocean water from the San Diego Mission Bay Jetty
 * capturing mice living in rural Arkansas

In order to perform an investigation you need something to investigate. OBI uses the standard label &quot;specimen creation&quot;, but in some cases &quot;specimen collection&quot; may be a more natural label (as the examples show).

Subject Enrollment

 * input: unspecified (outside the scope of the OBI model of investigations)
 * output: organism -- such as a human
 * examples:
 * human subject enrollment such as
 * enlisting family members of HIV patients into a study

Enrollment is a special form of specimen creation where an organism such as a human being is recruited into the trial. Enrollment may involve other steps such as informed consent process.

Material Processing

 * input: material entity -- such as a protein, cell, organism, or population
 * output: material entity
 * examples:
 * establishing cell culture
 * administering a substance in vivo such as
 * injecting mice with 10 ug morphine intranasally
 * a patient taking two pills of 1 mg aspirin orally

Once the specimens have been created they may require some material processing in order to prepare them for the assay.

Subject Processing

 * input: organism -- such as a human
 * output: organism
 * examples:
 * group assignment
 * study intervention

#|Subject processing is a special case of material processing involving organisms such as humans.

Assay

 * input: material entity -- such as a protein, cell, organism, or population
 * output: measurement datum
 * examples:
 * age measurement assay
 * DNA sequencing

An assay takes specimens or subjects and produces data about them. This is the most important step, where we switch from dealing with materials to dealing with information.

Data Processing

 * input: data item -- such as a measurement datum or data set
 * output: data item
 * examples:
 * data transformation such as
 * normalization data transformation
 * data combination
 * data interpretation such as
 * performing a diagnosis
 * data imputation
 * chi square test

Data transformation changes the data from one form to another without making further assumptions or interpretations. Data interpretation means stepping beyond the data you collected by interpolating, extrapolating, or making substantial assumptions.

Planning and Execution
The study design consists of one or more protocols. Each protocol consists of one or more objective specifications. An objective specification is a plan to execute some process. There are objective specifications for each of the processes discussed above:


 * 1) specimen creation objective and subject #|enrollment objective
 * 2) material transformation objective and #|subject processing objective
 * 3) assay objective
 * 4) #|data processing objective

In the study design execution stage the planned processes are executed. This is how OBI distinguishes between the plan to perform a process and the process itself.

Reporting
An #|investigation report is a special sort of document that describes the study design, the study design execution, the data collected, and the conclusions of the investigation. It is the sort of document that is regularly published in scientific journals.

= Development Notes =

The model of investigations described here does not match the current version of OBI. In order to make OBI conform to this model the following changes will be required:


 * 1) Add 'investigation report' as a child of report. It can have several parts:
 * 2) * study design
 * 3) * data item
 * 4) * 'conclusion textual entity'
 * 5) * perhaps 'hypothesis textual entity' -- modelling hypothesis may be too difficult
 * 6) Add 'investigation reporting' as a child of documenting.
 * 7) Modify investigation:
 * 8) * replace 'documenting' with 'investigation reporting'
 * 9) * replace 'conclusion textual entity' with 'investigation report' as an output
 * 10) Change material processing to have output material entity -- processed material is too specific.
 * 11) Move enrollment under specimen creation.
 * 12) Add 'subject processing' as a child of material processing.
 * 13) Change the output of assay to measurement datum.
 * 14) Rename 'interpreting data' to 'data interpretation'.
 * 15) Change the output of 'data interpretation' to data item.
 * 16) * This allows for a series of transformation and interpretation steps.
 * 17) * We should understand &quot;interpretation&quot; to mean an educated inference based on knowledge of the field. As such it is intended to produce truthful statements and should tend to do so reliably.
 * 18) * Understood in this way, the conclusion of an inductive inference fits the definition: &quot;a data item is an information content entity that is intended to be a truthful statement about something (modulo, e.g., measurement precision or other systematic errors) and is constructed/acquired by a method which reliably tends to produce (approximately) truthful statements.&quot;
 * 19) * 'Conclusion textual entity' should be reserved for labelling parts of documents. &quot;Conclusion&quot; is a role rather than a type of information content entity.
 * 20) Add 'data processing' as a child of planned process and
 * 21) * move data transformation under 'data processing'
 * 22) * move 'data interpretation' and under 'data processing'
 * 23) * move some of the current children of data transformation to 'data interpretation' -- anything involving interpolation, extrapolation, or inductive inference. For instance:
 * 24) ** background correction data transformation
 * 25) ** class discovery data transformation
 * 26) ** class prediction data transformation
 * 27) ** curve fitting data transformation
 * 28) ** data imputation
 * 29) ** performing a diagnosis
 * 30) ** error correction data transformation
 * 31) ** more ...
 * 32) Add under objective specification:
 * 33) * 'enrollment objective'
 * 34) * 'subject processing objective'
 * 35) * 'data processing objective'
 * 36) Move 'data transformation objective' under 'data processing objective'.
 * 37) Add 'data interpretation objective' under 'data processing objective'.

Other Concerns

 * It seems too strong to insist that a report is always the output of an investigation. Christian Bölling points out that there are many other possible outputs ("the study design itself, specimens, further processed specimens, data items of various complexity") and suggests that by modelling data processing and analysis more completely we can better capture the nature of a report as the output of an investigation.
 * It would be better to define some sort of inputs for specimen creation and subject enrollment.
 * Christian Bölling suggests that 'objective specification' should be renamed 'objective' on the following grounds: "it's the information content that matters and in this respect an objective is different from its specification. Of course, this is true also for the 'plan specification'. 'Plan' however, is already being used as label for specific, concretized 'plan specifications', so this would need to be renamed as well..."
 * 'subject enrollment' might be a better label for enrollment
 * 'material processing objective' might be a better label for material transformation objective
 * publication is not a type of document but a status (role?)
 * biological feature identification objective and children have no corresponding planned processes
 * group assignment should be a planned process