Methods definitions for CoMeth objects

[PeteHaitch]

Although CoMeth objects are based on SummarizedExperiment objects not all methods are identical between the two classes. This issue tracks three things:

1. Which methods have been implemented for CoMeth objects.
2. Whether these methods definitions are unique for CoMeth objects or inherited from other classes.
3. Any methods a user might reasonably expect to be implemented for the CoMeth class that I have chosen not to implement.

A tick means the method has been implemented, either specifically for CoMeth objects or inherited from another class. A method is not considered to be implemented unless unit tests have also been written.

[PeteHaitch]

Methods for CoMeth objects to classify

From SummarizedExperiment

• [ ] colData<-: Should check that column names/types are valid
• [ ] assay<-
• [ ] countOverlaps
• [ ] disjointBins
• [ ] distance
• [ ] distanceToNearest
• [ ] findOverlaps
• [ ] follow
• [ ] isDisjoint
• [ ] match
• [ ] nearest
• [ ] order
• [ ] overlapsAny
• [ ] precede
• [ ] ranges
• [ ] ranges<-
• [ ] rank
• [ ] rowData
• [ ] rowData<-
• [ ] seqinfo
• [ ] seqinfo<-
• [ ] seqnames
• [ ] sort
• [ ] split
• [ ] splitAsListReturnedClass
• [ ] start
• [ ] strand
• [ ] strand<-
• [ ] subsetByOverlaps
• [ ] width

From BiocGenerics

• [ ] append
• [ ] as.data.frame
• [ ] as.vector
• [ ] duplicated, anyDuplicated
• [ ] eval
• [ ] pmax, pmax.int
• [ ] pmin, pmin.int
• [ ] Reduce, Filter, Map, Position
• [ ] get, mget
• [ ] lapply, sapply
• [ ] mapply
• [ ] match
• [ ] paste
• [ ] rank
• [ ] rep.int
• [ ] rownames
• [ ] colnames
• [ ] union, intersect, setdiff
• [ ] table
• [ ] tapply
• [ ] unique
• [ ] unlist
• [ ] annotation, annotation<-
• [ ] normalize
• [ ] strand, strand<-
• [ ] updateObject

[PeteHaitch]

Methods defined for CoMeth object that are inherited from SummarizedExperiment

A tick means that the method has been tested interactively and unit test(s) written.

• [x] granges
• [ ] assay
• [x] colData
• [x] "["
• [x] dim
• [x] dimnames
• [x] end
• [ ] exptData: No unit test because I am not currently using this slot.
• [ ] exptData<-: No unit test because I am not currently using this slot.
• [x] mcols

[PeteHaitch]

Methods defined for CoMeth that are different to their definition for a SummarizedExperiment

A tick means the method has been implemented for CoMeth objects. The notes briefly describe the implementation.

• [x] cbind: cbind allows for the CoMeth objects to have different 'pos' data provided all objects have the same seqinfo. This is unlike cbind for SummarizedExperiment object, which requires all objects to have the same pos data (technically, rowData). Samples without counts data for a particular m-tuple will have the corresponding value filled with NA. The m of all objects must be identical. methylation_type may differ between object but the CoMeth object will have the combined methylation_type, e.g. cbind-ing a "CG" with "CHG" will result in a new object with a "CG/CHG" methylation_type. sampleNames must be unique across all objects being combined. The way cbind is implemented means that it effectively extracts all the relevant information from the objects being combined and then create a new CoMeth object with the CoMeth constructor.
• [x] rbind: rbind combines objects with different m-tuples and the same subjects (sampleNames). The sampleNames must match or an error is thrown. Duplicate columns of colData must contain the same data. Can't simply use the rbind method defined for a SummarizedExperiment because I also need to check the "extra positions" before concluding that an m-tuple is unique.
• [x] compare, ==, <=, !=, >=, <, > : compare compares all positions of each m-tuple in a CoMeth object. If I simply used the method defined for SummarizedExperiment objects then it would only check pos1 and posm. ==, <=, !=, >=, < and > all inherited from compare, as advised in the help page for Vector-comparison
• [x] duplicated: Checks all positions in m-tuple and the strand before concluding a CoMeth object contains duplicate m-tuples. If I used the method defined for SummarizedExperiment, only the the rowData would be checked.
• [ ] mcols<-: Checks that none of the new metadata column names begin with pos. Awaiting reply from Bioconductor help on whether my implementation of this idea is valid.

[PeteHaitch]

Methods defined for CoMeth that do not exist for SummarizedExperiment

A tick means the method has been implemented for CoMeth objects. The notes briefly describe the implementation.

• [x] length: Number of m-tuples in CoMeth, i.e. nrow(CoMeth).
• [x] sampleNames, sampleNames<-: Based on Biobase::sampleNames
• [x] ncol, NCOL: Inherited from ANY. Returns the number of samples in the CoMeth object although this behaviour cannot be guaranteed because I don't understand the inheritance.
• [x] nrow, NROW: Inherited from ANY. Returns the number of m-tuples in the CoMeth object although this behaviour cannot be guaranteed because I don't understand the inheritance.

[PeteHaitch]

Methods that I will not implement for CoMeth objects

Notes describe reason.

• [ ] combine: Note defined for SummarizedExperiment. Furthermore, cbind and rbind cover the two use cases for combine

[PeteHaitch]

Methods that I'm not sure what to do with

• [ ] coverage: As it stands, coverage(CoMeth) effectively returns the number of times each methylation loci appears in an m-tuple. This might be useful in some circumstances but it is very different to what I think of as the coverage of an m-tuple. My coverage of an m-tuple is the number of reads that overlap the m-tuple for a given sample, i.e. the sequencing coverage of that m-tuple for a given sample.
• [ ] end<-, start<-, width<-: Don't really want to give the user the ability to tinker with m-tuple positions.