The Impact of RNA Extraction Method on Ribosomal-depleted RNA Sequencing (RNA-Seq) Experiments in a Translational Setting

Background Ribosome depleted RNA-seq is well suited for understanding coding- and noncoding transcript expression in clinical samples with diverse input RNA quality and mass. Current choices for RNA extraction include phase-separation (e.g., ThermoFisher TRIzol) approaches to column-based (e.g., Norgen Spin Columns) approaches. Each method, however, has been noted to have differing effects on the properties of the extracted nucleic acids, precluding the straightforward bioinformatic comparison of data generated by the two methods. Methods In this study, we present differences in data differing only in total RNA extraction approaches (phase separation vs. column) from clinical samples processed by the same tissue bank and sequencing facility personnel. We use data from seven samples with libraries generated from RNA extracted with both extraction methods to characterize the effect of 'differentially extracted' genes (DExGs) on transcriptomic profiles. We then use another set of 148 samples extracted with one or the other method to confirm the differences observed in the 7-sample dataset. Results The large number of samples employed in this study allows us to find significant differences in important quality control parameters as well as many DExGs between the two methods. We identify several biophysical properties which are potential drivers of these differences. Conclusions We show that partial correction of extraction method mediated differences can be achieved but full correction is not possible. These results lay the foundation for principled decision making of RNA extraction methods for use in clinical samples in a translational environment, particularly in light of recent reagent shortages during the COVID-19 pandemic. Overall design: We use data from seven samples with libraries generated from RNA extracted with both extraction methods to characterize the effect of 'differentially extracted' genes (DExGs) on transcriptomic profiles. We then use another set of 148 samples extracted with one or the other method to confirm the differences observed in the 7-sample dataset.