modularize code klustering and check split_tetra

[csmiguel]

KTU2/R/KTU2.R

Lines 112 to 124 in a3a5235

	if(class(repseq)=="list" & all(names(repseq)==c("tetra.table","output.seq"))){
	message("loading k-mer frequency table...")
	tetra.table <- repseq[[1]]
	asv.id <- colnames(tetra.table)
	species <- repseq[[2]]
	} else{
	message("k-mer frequency calling...")
	tetra.table <- tetra.freq(repseq, pscore = as.logical(pscore), file = as.logical(seqfromfile), cores=cores)
	asv.id <- colnames(tetra.table)
	if(class(repseq)=="DNAStringSet"){
	species <- as.character(repseq)
	} else species <- as.character(Biostrings::readDNAStringSet(filepath = repseq,use.names = T))
	}

I would try to modularize the code as much as possible. For instance, in this case, I would force klustering to only accept tetra.tables or fasta files. This is a way to make the code cleaner by avoiding redundancies and it is easier to debug later.
My suggestion would be to let it feed only from tetra.tables.

On the same line of thinking I noticed the function split_tetra computes tetra.table and after splits the sequences based on these tables. To modularize the code I would take out from the function the lines computing tetra tables:

KTU2/R/KTU2.R

Line 516 in a3a5235

tetra <- tetra.freq(repseq,output.seq = TRUE, cores = cores)

Other enhancements regarding split_tetra:

1. returns an error when the length(tetra.table) < 2. This can happen for small datasets when there is no split. Please, try to reproduce the error.
2. it is time-consuming given the simplicity of the calculations. I don't know what step is bottlenecking the process. Other k-mer-based clustering algorithms might be faster. For instance: https://doi.org/10.1093/nar/gky315.