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

The translocation speed of the biological molecule through the pore is not homogenous. Thus, the translocation rate for each individual nucleotide can differ significantly. The translocation speed can be influenced by the sequencing context [31] , but also by sample time and sequencing buffer conditions (unpublished observations). Furthermore, the translocation speed can be influenced by RNA or DNA modifications, which however should not affect RNA sequencing from in vitro transcribed molecules, or DNA sequencing PCR-amplified molecules. Importantly, the motor protein for RNA and DNA differs, leading to dramatically different average translocation speed (70 nt for RNA in ONT Kits SQK-RNA001 vs. 450 nt for DNA in ONT Kits SQK-LSK108). In conclusion it is important to estimate the average . 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 nucleotide translocation rate for each read separately to account for the specific conditions at which the read was recorded. In basecalled ONT sequence data, a 'move' in raw data describes a single-nucleotide translocation through the pore. To calculate the average read-specific nucleotide translocation rate, we first extract a vector containing the number of sample points per move from the FAST5 events table of each individual read. If a move of 2 is detected (does not occur in basecalling with Flip-flop models), we divide the number of sample points by 2, as we reason that a preceding nucleotide translocation was not detected by the basecaller. From the resulting distribution of sample points per move, we then compute the geometric mean. This strategy results in robust estimation of poly(A) tail length for both ONT RNA and DNA sequencing approaches with standard model basecalling ( Fig. 1D; Fig. 2C ). In our experience, this approach is resulting in more robust normalisation compared to normalisation by the median of single-nucleotide translocation rates, which is used in Nanopolish ( Fig. S2B ; [32] ).