filter_regions: Filtering process prior to running Melissa
In Melissa: Bayesian clustering and imputationa of single cell methylomes

Description Usage Arguments Details Value Author(s) See Also Examples

Fuctions for filter genomic regions due to (1) low CpG coverage, (2) low coverage across cells, or (3) low mean methylation variability.

filter_by_cpg_coverage(obj, min_cpgcov = 10)

filter_by_coverage_across_cells(obj, min_cell_cov_prcg = 0.5)

filter_by_variability(obj, min_var = 0.1)

`obj`	Melissa data object.
`min_cpgcov`	Minimum CpG coverage for each genomic region.
`min_cell_cov_prcg`	Threshold on the proportion of cells that have coverage for each region.
`min_var`	Minimum variability of mean methylation across cells, measured in terms of standard deviation.

The (1) 'filter_by_cpg_coverage' function does not actually remove the region, it only sets NA to those regions. The (2) 'filter_by_coverage_across_cells' function keeps regions from which we can share information across cells. The (3) 'filter_by_variability' function keeps variable regions which are informative for cell subtype identification.

The filtered Melissa data object

C.A.Kapourani C.A.Kapourani@ed.ac.uk

melissa, create_melissa_data_obj

# Run on synthetic data from Melissa package
filt_obj <- filter_by_cpg_coverage(melissa_encode_dt, min_cpgcov = 20)

# Run on synthetic data from Melissa package
filt_obj <- filter_by_coverage_across_cells(melissa_encode_dt,
                                            min_cell_cov_prcg = 0.7)

# Run on synthetic data from Melissa package
filt_obj <- filter_by_variability(melissa_encode_dt, min_var = 0.1)