Reads were subsampled at 1, 2, 4, 8, and 16 x 100k pairs from trimmed, normalised, paired 100bp rice Illumina reads. Assemblies were performed using default settings for SOAPDenovoTrans (soapdt) with varying values of k from 21-86 with step 4. All calculations were run on a 24-core Intel i7 node with 100GB RAM and with files stored in tmpfs (ramdisk) for speed.
Notably, the full k-sweep with 100k reads took only around one minute.
Three objective functions were used to analyse each assembly:
