PAMES has been completely rewritten to work with our new framework MiMeSis Major changes:
- all functions now return a data.frame instead of a vector: according to our experience, data.frames integrate better with pipelines;
get_AUCreplacescompute_AUCand takes as input a matrix of beta-values as originally produced by Illumina Beadchips (instead of rounded percentages);find_informative_sitesandfind_informative_regionsreplaceselect_informative_sitesandselect_informative_regions, respectively, and integrate the newcontrol_constraintsparameter;get_purityreplacecompute_purityand takes as input a data.frame (generated byfind_informative_sites/find_informative_regions).
- rename
median_of_regions:reduce_region - add
methodparameter toreduce_to_regions: allow choice between "median" (default) or mean
- fix
select_informative_regions_extwith flagreturn_info=TRUE
- renamed parameter
platformtoref_table
- added new function (select_informative_sites_ext.R) that:
- considers also control data distribution
- use custom platform data (useful to filter out chromosomes X and Y)
- return named vectors (names are either derived from rownames of tumor_data or created with format "CpG_xxxxxxx")
- compute_purity now requires platform data (not just "450k" or "27k") addressing issue#7)
- deprecated parameter na_threshold in favor of min_sample_frac
compute_AUC - added simplify parameter: if false return a data.frame reporting fraction of NA samples per site
compute_AUCnow returns a vector with the names of the CpG probes.
- Add
platformparameter tocompute_purityto address issue #7)
- Fix bug in
reduce_to_regions(see issue #6)
- Fix bug in
compute_AUC
- Add
percentilesparameter for a more flexible selection of sites
- Even faster cluster reduction
- Fix bug: "top" method retrieved same sites of "hyper"
- Faster and simpler cluster reduction in
select_informative_sites - Removed internal functions:
is_too_closecluster_reduction
- Added variable
methodtoselect_informative_sitesandselect_informative_regionsto select even, top, hyper or hypo methylated sites.
-
PAMES have been rewritten to make the code simpler and the analysis run faster. Thanks to GMFranceschini for suggesting to use
GenomicRanges. -
Removed function:
compute_islands_indexes.R
-
Renamed functions:
reduce_to_islands.R->reduce_to_regions.R
-
Renamed data:
cpg_islands_df->cpg_islands
-
Removed data:
bs_toy_indexes
- BUG correction: setting na_threshold to be greater than 1 didn't generate an error (and resulted as having na_threshold set to 0.99)
- Faster compute_AUC thanks to Wilcoxon method
- Added parallel computation to
compute_AUCwith argumentncores.
cpg_islandshas been renamed ascpg_islands_dfbs_tumor_toy_dataandbs_control_toy_datahas been merged intobs_toy_matrix
- All beta tables are now required to be percentage values.