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.peakMode`` Selects ROCCO export shape. ``narrow`` Writes `UCSC narrowPeak `_ calls. ``broad`` Writes `UCSC gappedPeak `_ calls. ``both`` Writes narrow and broad calls. This is the default. ``matchingParams.thresholdZ`` Sets a one-sided :math:`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.minPeakScore`` Filter selected peaks to those with an average signal above this threshold. Uncertainty Score """"""""""""""""" ``matchingParams.uncertaintyScoreMode`` ``state`` uses the fitted state track directly. ``lower_confidence`` uses ``state - matchingParams.uncertaintyScoreZ * uncertainty`` to penalize regions where estimates are uncertain. ``matchingParams.uncertaintyScoreZ`` Sets the multiplier used by ``lower_confidence`` scoring. Larger values penalize uncertain regions more strongly. Estimation Controls ~~~~~~~~~~~~~~~~~~~ ``countingParams.intervalSizeBP`` Sets 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. ``fitParams.ECM_backgroundLengthScaleMultiplier`` Sets the multiplier that converts the inferred correlation-length into the soft background fitting window. Larger values softly restrict the shared background estimate :math:`g_{[i=1,\ldots,i=n]}` to lower frequencies. ``outputParams.stateShrinkageSpikeOddsMultiplier`` (Experimental) Ignored if posterior state shrinkage is disabled entirely (``outputParams.stateShrinkageEnabled``). Values above `1.0` multiply fitted point-null odds upward, which can reduce false positives in low-signal regions. Values below `1.0` multiply those odds downward.