tailfindr: Alignment-free poly(A) length measurement for Oxford Nanopore RNA and DNA sequencing

Maximilian Krause; Adnan M. Niazi; Kornel Labun; Yamila N. Torres Cleuren; Florian S. Müller; Eivind Valen

We further show that tailfindr poly(A) tail estimates agree closely with a recently developed tool that relies on prior mapping of the data [32] . While poly(A) tail boundaries in raw signal are found to be essentially the same with the two different approaches (Fig. S2C,D) , the final calculated poly(A) tail lengths differ (Fig. S2A) . Specifically, tailfindr estimates short poly(A) stretches slightly longer than Nanopolish, while long poly(A) stretches result in shorter estimates in tailfindr. These differences can be explained by different calculation of the average nucleotide translocation rate ( Fig. S2B) which is used to normalise raw poly(A) tail measurements. Nanopolish normalises by calculating the readspecific median of the samples per nucleotide after removing 5% of the translocation rate outliers. We have observed that this normalisation is resulting in correct poly(A) estimation in RNA, but not DNA sequencing approaches. . CC-BY-NC-ND 4.0 International license peer-reviewed) is the author/funder. It is made available under a The copyright holder for this preprint (which was not . https://doi.org/10.1101/588343 doi: bioRxiv preprint Instead we normalise by the read-specific geometric mean of samples per nucleotide without outlier removal. Another difference between the tools is that tailfindr does not need any sequence data preprocessing, as it only requires basecalled FAST5 files as input. This allows for poly(A) tail studies independent of any other tool than the essential basecaller, which would allow for an integration of the tailfindr algorithm into the basecalling procedure. This in turn makes it possible to assign poly(A) tail lengths to individual reads in parallel to the sequencing procedure, making live poly(A) tail analysis feasible.