Use this function to find all biomarkers across multiple performance classification group matchings based on a given threshold between 0 and 1.
get_biomarkers(diff.mat, threshold)
| diff.mat | a matrix whose rows are vectors of average node data
differences between two groups of models based on some kind of classification
(e.g. number of TP predictions) and whose names are set in the |
|---|---|
| threshold | numeric. A number in the [0,1] interval, above which (or below its negative value) a biomarker will be registered in the returned result. Values closer to 1 translate to a more strict threshold and thus less biomarkers are found. |
a list with two elements:
biomarkers.pos: a character vector that includes the node
names of the positive biomarkers
biomarkers.neg: a character vector that includes the node
names of the negative biomarkers
This function uses the get_biomarkers_per_type function
to get the biomarkers (nodes) of both types (positive and negative) from the
average data differences matrix. The logic behind the biomarker selection is
that if there is at least one value in a column of the diff.mat matrix
that surpasses the threshold given, then the corresponding node (name of the
column) is returned as a biomarker.
This means that for a single node, if at least one value that represents an average data
difference (for example, the average activity state difference) between any
of the given classification group comparisons is above the given threshold
(or below the negative symmetric threshold), then a positive
(negative) biomarker is reported.
In the case of a node which is found to surpass the significance threshold level given both negatively and positively, we will keep it as a biomarker in the category which corresponds to the comparison of the highest classification groups. For example, if the data comes from a model performance classification based on the MCC score and in the comparison of the MCC classes (1,3) the node of interest had an average difference of -0.89 (a negative biomarker) while for the comparison of the (3,4) MCC classes it had a value of 0.91 (a positive biomarker), then we will keep that node only as a positive biomarker. The logic behind this is that the 'higher' performance-wise are the classification groups that we compare, the more sure we are that the average data difference corresponds to a better indicator for the type of the biomarker found.
Other biomarker functions:
get_biomarkers_per_type()