Signal aggregation functions

Signal aggregation functions for circularProjection and polarProjection internal calls. The aggregation should be symmetric with respect to signal polarity, ensuring that opposite signals produce corresponding outputs.

Usage

signalAggregation(method = c("mean", "wmean", "log.wmean", "exp.wmean"))

Arguments

method: A character string specifying the method for signal aggregation, returning either a customized mean or weighted.mean function.

Value

Returns a function of the form: function(x) { ... }

Details

At each point in pathway space, multiple vertices may each contribute a decayed signal value; method controls how these contributions are combined into a single value:

mean: a plain, unweighted average. Because the same number of potential contributors is assumed throughout the image, points reached by few vertices can show a diluted value compared to points reached by many, even when the underlying signals are equally strong.
wmean: each contribution is weighted by its own magnitude, so the strongest nearby signal dominates the result regardless of how many (or how weak) the other contributions are.
log.wmean: like wmean, but the weighting is compressed, giving moderate signals comparatively more influence relative to the single strongest one.
exp.wmean: like wmean, but the weighting is sharpened, so the strongest nearby signal dominates the result even more than under wmean.

Unlike mean, the weighted variants are not affected by the dilution described above, since contributions with zero weight do not affect their result. mean is a reasonable default for simple or binary signals, such as those used in introductory examples; for continuous, more nuanced analyses, wmean (or one of its variants) is generally preferable. For other aggregation rules, see fuzzy logic functions in the online tutorials: https://sysbiolab.github.io/PathwaySpace/

Author

Sysbiolab Team

Examples

aggregate.fun <- signalAggregation()