MO Astray V1 0 5-SiMPLEX
Considering only the results on simulated data, one might prefer quantification based on alignment or lightweight mapping of sequencing reads directly to the transcriptome, rather than performing alignment to the genome followed by projection to the transcriptome. One would also observe only small differences between lightweight mapping and alignment to the transcriptome. However, our analyses in experimental data suggested that the increased complexity in real RNA-seq experiments leads to more divergent behavior. In both the bulk and full-length single-cell samples analyzed, SAF yielded the highest overall correlation with the oracle, despite the fact that the oracle is derived from a combination of the Bowtie2 and STAR alignment results. Among the methods based on traditional alignment, alignment to the genome (using STAR, and projecting the resulting alignments to the transcriptome) seemed to display the best concordance, on average, with the quantifications resulting from oracle alignments. SA yielded similar but slightly better accuracy than alignment to the transcriptome using Bowtie2. This is likely, in part, because it is accounting for the sequence similar decoys that can lead alignment to only the target transcriptome astray. The main benefit of SAF is that it aligns to a reference index that contains both the fully spliced transcript sequences as well as the entire underlying genome (as potential decoy sequence). This allows SAF to obtain the type of sensitivity that is exhibited by approaches like Bowtie2 and SA when the read truly arises from the annotated transcriptome, but also allows it, like STAR, to avoid spuriously aligning a read to an annotated transcript when it is better explained by some other genomic locus. In the experimental data, both alignment-based approaches and selective alignment methodologies performed better than quasi-mapping, though the manner in which these methods differ from quasi-mapping, and from each other, was not identical.
MO Astray V1 0 5-SiMPLEX