In the absence of a reference genome, the ultimate goal of a de novo transcriptome assembly is to accurately and comprehensively reconstruct the set of messenger RNA transcripts represented in the sample. Nonreference assembly of the transcriptome of polyploid species poses a particular challenge because of the presence of homeologs that are difficult to disentangle at the sequence level. This is especially true for hexaploid oats, which have three highly similar subgenomes, two of which are thought to be nearly identical. Under these circumstances, most software packages and established pipelines encounter difficulties in rendering an accurate transcriptome because they are typically developed, refined, and tested for diploid organisms. We present a protocol for transcriptome assembly in oats that can be extended both to other polyploids and species with highly duplicated genomes.