For a given stringency level, the presence of a nucleosome is inferred, if the enrichment value is ≥ the value of the stringency level continuously at 4 probes (for distinct nucleosomes) or at >4 probes (for fuzzy nucleosomes) in all the three replicates in the dataset under study.
At this stage of the analysis, E-values of low stringency were used for BLAST and HMM search algorithms to enable detection of all true positives.
Large values of T (high stringency) is appropriate for gene expression [ 15], while small values of T (low stringency) increase the weighting of subtle differences in expression, and may be required to discriminate arrays at the pathway level, where the coordinated effects of multiple genes are under consideration.
Higher values of β increase statistical stringency by placing lower emphasis on smaller correlation differences.
Efficiency of the method enables the analysis to be performed for a range δ and ζ values that correspond to different levels of stringency.
For the Miso method, the tuning parameter is a p-value cutoff and therefore is a measure of stringency (i.e. the closer the parameter is to zero the higher the positive predictive value).
HMM searches were performed, as previously described, in a series of threshold E values, increasing stringency from E value < 0.5 to E value < 0.05 (data not shown).
The level of stringency applied when determining the cutoff value influences the prevalence rates.
Its natural interpretation as a measure of stringency provides external guidance when selecting a value appropriate for a specific application and, more generally, implies a predictable relationship between its value and classical measures of performance.
We applied this tool using two levels of stringency, medium and highest, and recorded the level of significance (the Geo-value) and the terms shared.
