GLM_inference: Statistical Inference with DESeq2 on IP over input fold...
In exomePeak2: Bias Awared Peak Calling and Quantification for MeRIP-Seq

GLM_inference conduct inference on log2 fold changes of IP over input using the GLM defined in DESeq2.

GLM_inference(
  SE_bins,
  glm_type = c("Poisson", "NB", "DESeq2"),
  p_cutoff = 1e-05,
  p_adj_cutoff = NULL,
  count_cutoff = 5,
  log2FC_mod = 1,
  min_mod_number = NA,
  correct_GC_bg = FALSE,
  qtnorm = TRUE,
  consistent_peak = FALSE,
  consistent_log2FC_cutoff = 1,
  consistent_fdr_cutoff = 0.05,
  alpha = 0.05,
  p0 = 0.8
)

`SE_bins`	a `SummarizedExperiment` of read count. It should contains a colData with column named design_IP, which is a character vector with values of "IP" and "input". The column helps to index the design of MeRIP-Seq experiment.
`glm_type`	a `character`, which can be one of the "Poisson", "NB", and "DESeq2". This argument specify the type of generalized linear model used in peak calling; Default to be "Poisson". The DESeq2 method is only recommended for high power experiments with more than 3 biological replicates for both IP and input.
`p_cutoff`	a `numeric` for the p value cutoff used in DESeq inference.
`p_adj_cutoff`	a `numeric` for the adjusted p value cutoff used in DESeq2 inference; if provided, values in `p_cutoff` will be ignored.
`count_cutoff`	an `integer` indicating the cutoff of the mean of reads count in a row, inference is only performed on the windows with read count bigger than the cutoff. Default value is 10.
`log2FC_mod`	a non negative `numeric` for the log2 fold change cutoff used in DESeq inferene for modification containing peaks (IP > input).
`min_mod_number`	a non negative `numeric` for the minimum number of the reported modification containing bins. If the bins are filtered less than this number by the p values or effect sizes, more sites will be reported by the order of the p value until it reaches this number; Default to be calculated by floor( sum(rowSums( assay(SE_bins) ) > 0)*0.001 ).
`correct_GC_bg`	a `logical` of whether to estimate the GC content linear effect on background regions; default `= FALSE`. If `correct_GC_bg = TRUE`, it may result in a more accurate estimation of the technical effect of GC content for the RNA modifications that are highly biologically related to GC content.
`qtnorm`	a `logical` of whether to perform subset quantile normalization after the GC content linear effect correction; default `= TRUE`. Subset quantile normalization will be applied within the IP and input samples seperately to account for the inherent differences between the marginal distributions of IP and input samples.
`consistent_peak`	a `logical` of whether the positive peaks returned should be consistent among replicates; default `= TRUE`.
`consistent_log2FC_cutoff`	a `numeric` for the modification log2 fold changes cutoff in the peak consisency calculation; default = 1.
`consistent_fdr_cutoff`	a `numeric` for the BH adjusted C-test p values cutoff in the peak consistency calculation; default = 0.05. Check `ctest`.
`alpha`	a `numeric` for the binomial quantile used in the consitent peak filter; default `= 0.05`.
`p0`	a `numeric` for the binomial proportion parameter used in the consistent peak filter; default `= 0.8`. For a peak to be consistently methylated, the minimum number of significant enriched replicate pairs is defined as the 1 - alpha quantile of a binomial distribution with p = p0 and N = number of possible pairs between replicates. The consistency defined in this way is equivalent to the rejection of an exact binomial test with null hypothesis of p < p0 and N = replicates number of IP * replicates number of input.