Parameter Guidance¶
Most parameters can be left at their default values. The following sections provide guidance on key arguments that may be useful to adjust for specific applications.
Peak Calling Controls¶
matchingParams.peakModeSpecifies calling of narrow and/or broad marks.
narrowWrites UCSC narrowPeak calls.
broadWrites UCSC gappedPeak calls.
bothWrites narrow and broad calls.
narrowis the default.matchingParams.thresholdZSets a one-sided \(z\)-score cutoff used to define the budget (max. proportion of genome called). Larger values will yield a smaller budget and fewer calls. Smaller values will yield a larger budget and more calls. The default value is 2.0.
matchingParams.minMeanSignalKeeps regions whose covered-BP-weighted mean signal reaches the supplied descriptive threshold. Disable with
null.matchingParams.mergeToleranceBPSets the maximum gap eligible for broad-family merging. A positive value is required for broad and combined modes.
matchingParams.maxRegionBPSets the maximum outer width for a broad family. A positive value is required for broad and combined modes.
Estimation Controls¶
countingParams.intervalSizeBPSets the genomic bin size in base pairs. The default is 50 and is appropriate for most cases. Higher-resolution results may be obtained using 25, 10, etc. For detecting domain-level enriched-regions in broad marks like H3K27me3, larger values (100, 250, etc.) should suffice.
observationParams.precisionMultiplierMinandobservationParams.precisionMultiplierMaxBound observation precision multipliers. The defaults are
0.1and10.0.processParams.precisionMultiplierMinandprocessParams.precisionMultiplierMaxBound process precision multipliers. The defaults are
1.0e-4and10.0. Because \(Q_i=Q_0/\kappa_i\), the multipliers let the local frequency response adapt: smaller \(\kappa_i\) admits faster variation, whereas larger \(\kappa_i\) favors smoother variation.outputParams.stateShrinkageEnabledEnables experimental posterior state shrinkage and defaults to
True. Set it toFalseto disable it. Use the shrunk state as a visualization or ranking track, rather than as a replacement for the fitted state.