I have paired-end Illumina reads, R1.fq and R2.fq. From these I merge overlapping reads, using BBmerge, and get the file merged.fq. In SPAdes I can input both paired-end (options --pe1-1 and --pe1-2) files and the merged.fq file (option --pe1-m). My question is: Should the paired-end files be the _original files_, containing all reads, or only the _leftover files_ with reads that did not merge? From what I read in the manual I would say the latter, but it is not entirely clear. One may ask why bother? Well, it might be a memory issue, since these samples have short insert size, and a large fraction of the reads merge .Thus, providing them twice (both in paired-end and merged files) seems like a waste of memory.