Open dlaehnemann opened 4 years ago
The implementation is fine. I am just wondering if this is really what one wants to do. If units are just unmerged lanes, then yes. If they are technical replicates, then probably no. Let us have a chat about this when I am in the office on Thursday.
Sure, let's discuss further. My first instinct would be for technical replicates to be modelled via the batch variables...
This PR implements sample merging by providing kallisto with all the units of a sample at once. I.e. if multiple sequencing runs are provided for a sample as units, kallisto will jointly quantify abundance for all the runs. In addition, the workflow now checks whether units of a sample are either all paired end or all single end and throws an exception if this is violated for a sample.