Limitations on the number of samples.

[kasperdanielhansen]

Currently we have a limitation on the number of samples supported by bsseq, based on R limitations on the size of a matrix. This translates into ~85 samples with 26M CpGs. We need to address this.

[kasperdanielhansen]

Possible solution 1: If R supports long vectors in matrices, we are good.
Possible solution 2a: Invent a new class which is essentially a data.frame-matrix, ie. a list of equally long vectors with the same storage mode. And then implement matrixStats on this.
Possible solution 2b: Store methylation as a list (per chromosome) of matrices and not a single big matrix.
Possible solution 3. Use on disk storage and use data representation which prevents loading everything into memory.