drio
commented
10 years ago After reading cubism's internals I think it makes more sense if we start from scratch and incorporate the different cubism components one by one, making the necessary changes to reflect that we are working with dna instead of time series. Also, we want to change the semantics to match the new domain (dna instead of time).
I have a prototype working but I think I am going to go ahead and create my own library (based on cubism but for dna). Not only everything is tied to time (obviously, after all this is a time series library) but plenty of logic can be removed if using working with plain non time values.
Now that I think about it, the new library should not be dna specific but generic. Data will be just numbers. In that way, people can create their own contexts and metrics. For example, for dna we would have: context: local_simple_bed and each metric will be a sample name. The machinery behind the context will pull data from name.bed, extract the data and return the array of values that cubism expects.
Promising.
cubism works only for time series data. Porting cubism to work with DNA data (bed files), would be beneficial for the community. In the dna branch you can find my attempts on accomplishing this. I will keep the progress on this ticket. Hacking time!