Preparation steps for RNA-Seq lead to both positional bias, whereby fragments are preferential located towards either the beginning or end of transcripts, and sequence-specific bias, whereby the sequence surrounding the beginning or end of potential fragments affects their likelihood of being selected for sequencing, in sequenced fragments. If such biases are not corrected for, expression estimates may lead to over or under-representation of fragments, thereby confounding differential expression and other related analysis.
In this paper, Roberts et al. estimate the extent and nature of these biases using a likelihood based approach. They find that bias correction improves the correlation of expression estimated obtained from sequencing data generated using different sample preparations and sequencing technologies. The software used has been packaged and incorporated into the Cufflinks RNA-Seq analysis suite.