This documentation describes the general concepts of feature grouping, which can be achieved by the different approaches described further below.
The main function for the stepwise feature grouping is groupFeatures. The
selection of the actual grouping algorithm (along with the definition of its
parameters) is done by passing the respective parameter object, along with
the object containing the input data and optional additional arguments,
to the groupFeatures method.
Arguments
- object
input data object on which (with which data) the feature grouping should be performed.
- param
parameter object which type defines the selection of the grouping algorithm.
- ...
additional arguments to be passed to the grouping algorithm.
Value
Depending on the implementation and the input object. Generally the input object with grouping results added. See respective help pages for more information.
Single-step Feature Grouping
Each feature grouping algorithm can be applied individually as a single-step approach, e.g. by grouping features only on a single feature property, such as the retention time. Additional feature grouping approaches might also be implemented that consider combination of different MS feature properties in a single clustering process.
Stepwise Feature Grouping Refinement
Stepwise feature grouping evaluates a single property of MS features (such
as their retention time or abundances) at a time to define the feature
groups. Each subsequent grouping step builds on the previous one by
eventually sub-grouping each feature group, if needed. Thus, feature groups
get refined in each step. As an example, grouping of features based on a
similar retention time would loosely group features from all compounds
eluting at about the same time from a e.g. liquid chromatography run. This
obviously would also group features representing ions from different
co-eluting compounds. Thus, calling groupFeatures on the previous feature
grouping result with a different parameter object would refine these
initial feature groups, splitting them based on another property of the
features (such as correlation of feature abundances across samples).
The advantage of the stepwise approach is that results can be evaluated after each grouping step and parameters adapted if needed. Also, it provides flexibility by allowing to change the order of grouping approaches, or skip individual steps if not suitable for the available data or the experimental setup.
The major disadvantage is that a wrong group assignment in one of the initial steps can not be corrected for in later steps.
See also
featureGroups() for the function to extract (defined) feature
groups from a result object.