A Cautionary Tale of Using RNAseq Data: Examples from Large-Scale Projects at the Umeå Plant Science Centre

RNASeq has rapidly replaced microarrays as the method of choice for  transcriptomics studies in both model and novel experimental systems. The rapid adoption of RNA-Seq has seen an explosion in the number of tools for performing all steps of the analysis pipeline. Although standards are slowly being established, as they were following the emergence of expression microarrays, the vast majority of existing tools has been developed using well assembled, properly annotated reference genomes. As such, some of these tools priors do not hold when working with non-model organisms and a detailed understanding of tools’ assumptions and data models is crucial to avoid severe pitfalls.

Across a number of large-scale RNA-Seq studies conducted at the Umeå Plant Science Centre in both Populus tremula and Picea abies, we have encountered several such issues. Mitigating them was neither technically nor computationally challenging, but was essential to obtaining reliable results. 

Through a series of examples, we will show:

• the importance of refining genome annotation: benefitting from known/novel splice junctions, adding novel genes and removing likely chimeras;
• how to increase the gene-space contiguity of novel genomes using RNA-Seq data;
• how these combined improvements helped decrease the false positive rate of differential expression analyses;

hence, telling a cautionary tale of how to use RNA-Seq as efficiently and accurately as possible, concluding by a display of how the knowledge gained is helping us further understand wood and leaf development and answering questions that eventually will benefit the forest industry.