tailfindr is a R package for estimating poly(A)-tail lengths in Oxford Nanopore reads.

• Works for both RNA and DNA reads. In the case of DNA reads, it estimates both poly(A)- and poly(T)-tail lengths.
• Supports data that has been basecalled with Albacore or Guppy. It also support data that has been basecalled using the newer ‘flipflop’ model.
• Can work on single or multi-fast5 file reads.

tailfindr has been developed at Valen Lab in Computational Biology Unit at the University of Bergen, Norway.

VBZ is a compression format used by Nanopore to compress the data inside a FAST5 file. You must install VBZ plugin for your operating system so that tailfindr can successfully read FAST5 files. Download and install VBZ plugin from here.

tailfindr depends on the HDF5 library for reading Fast5 files. For OS X and Linux, the HDF5 library needs to be installed via one of the (shell) commands specified below:

HDF5 1.8.14 has been pre-compiled for Windows and is available here — thus no manual installation is required.

Currently, tailfindr is not listed on CRAN/Bioconductor, so you need to install it using devtools. To install devtools use the following command in R/R-studio:

install.packages("devtools")

rbokeh has been delisted from CRAN and therefore tailfindr cannot install itself. You need to manually install it yourself using the command below:

devtools::install_url('https://cran.r-project.org/src/contrib/Archive/rbokeh/rbokeh_0.5.1.tar.gz', type = "source", dependencies = TRUE)

Now you can install tailfindr using the command below in R/R-studio:

devtools::install_github("adnaniazi/tailfindr")

If you also want to build the vignette while installing tailfindr, then run the command below:

remotes::install_github('adnaniazi/tailfindr', build = TRUE, build_opts = c("--no-resave-data", "--no-manual"), force = TRUE)

Now you are ready to use tailfindr.

find_tails() is the main function that you can use to find tail lengths in both RNA and DNA reads. It saves a CSV file containing all the tail-length data. Furthermore, it also returns the same data as a tibble.

Give below is a minimal use case in which we will run tailfindr on example RNA reads present in the tailfindr package.

library(tailfindr)
df <- find_tails(fast5_dir = system.file('extdata', 'rna', package = 'tailfindr'),
                 save_dir = '~/Downloads',
                 csv_filename = 'rna_tails.csv',
                 num_cores = 2)

In the above example, tailfindr returns a tibble containing the tail data which is then stored in the variable df. tailfindr also saves this dataframe as a csv file (rna_tails.csv) in the user-specified save_dir, which in this case is set to ~/Downloads. A logfile is also saved in the save_dir. The parameter num_cores can be increased depending on the number of physical cores at your disposal.

Additionally, tailfindr allows you to generate plots that show the tail location in the raw squiggle. You can save these plots as interactive .html files by using 'rbokeh' as the plotting_library. You can zoom in on the tail region in the squiggle and see the exact location of the tail.

Give below is a minimal use case in which we will run tailfindr on example cDNA reads present in the tailfindr package, and also save the plots:

df <- find_tails(fast5_dir = system.file('extdata', 'cdna', package = 'tailfindr'),
                 save_dir = '~/Downloads',
                 csv_filename = 'cdna_tails.csv',
                 num_cores = 2,
                 save_plots = TRUE,
                 plotting_library = 'rbokeh')

However, note that generating plots can slow down the performance of tailfindr. We recommend that you generate these plots only for a small subset of your reads.

tailfindr can plot additional information that it used while deriving the tail boundaries. Please read our preprint to learn how tailfindr works. To plot this information, set the plot_debug_traces parameter to TRUE.

df <- find_tails(fast5_dir = system.file('extdata', 'cdna', package = 'tailfindr'),
                 save_dir = '~/Downloads',
                 csv_filename = 'cdna_tails.csv',
                 num_cores = 2,
                 save_plots = TRUE,
                 plot_debug_traces = TRUE,
                 plotting_library = 'rbokeh')

tailfindr needs Fastq and Events/Move table to work on. By default, it searches for them in the Basecall_1D_000 group in the Analyses section of the FAST5 file. If for whatever reason, you need tailfindr to read data from another basecall group – lets say Basecall_1D_001 – then you can run tailfindr as below:

df <- find_tails(fast5_dir = system.file('extdata', 'rna_basecall_1D_001', package = 'tailfindr'),
                 save_dir = '~/Downloads',
                 csv_filename = 'rna_tails.csv',
                 num_cores = 2,
                 basecall_group = 'Basecall_1D_001',
                 save_plots = TRUE,
                 plot_debug_traces = TRUE,
                 plotting_library = 'rbokeh')

In this case, the input FAST5 have two basecall groups: Basecall_1D_000 and Basecall_1D_001 but we configured tailfindr to use Events/Move table from the Basecall_1D_001 group.

There are more options available in the find_tails() function. Please see its documentation.

If you have used custom front and end primers while designing you cDNA sequences, you can now specify them in tailfindr. tailfindr will use these sequences instead of the defaults ones to find the orientation of the reads and one of the ends of the poly(A)/(T) tail. Here is how you call tailfindr if you have used custom primers:

df <- find_tails(fast5_dir = system.file('extdata', 'cdna', package = 'tailfindr'),
                 save_dir = '~/Downloads/tailfindr_output',
                 csv_filename = 'cdna_tails.csv',
                 num_cores = 4,
                 dna_datatype = 'custom-cdna',
                 front_primer = "TTTCTGTTGGTGCTGATATTGCTGCCATTACGGCCGG",
                 end_primer = "ACTTGCCTGTCGCTCTATCTT")

Important thing to note here is the use of three additional parameters: dna_datatype, front_primer, and end_primer.

front_primer and end_primer sequences should always be specified in the 5’ to 3’ direction.

tailfindr returns tail data in a dataframe and also saves this information in a user-specified CSV file. The columns generated depend on the whether tailfindr was run on RNA or DNA data. Below is a description of columns for both thses scenarios:

Here are the columns that you will get from tailfindr if you have run it on DNA data:

• tailfindr needs the Events/Move table in the FAST5 file to calculate the read-specific normalizer – samples_per_nt – which is used to convert tail length in samples to tail length in nucleotides. If your data was basecalled with MinKNOW-Live-Basecalling, then the Events/Move table might not be saved in the FAST5 file. In such a case, you can rebasecall your reads and adjust the basecall_group parameter accordingly when calling find_tails() function as demonstrated in the use case # 4 above. This is because now the Events/Move table will now be under Basecall_1D_001 instead of tailfindr’s default search location Basecall_1D_000. See the figure below: The panel on left shows that the MinKNOW live basecalled read; it has no Event/Move table. The panel on the right shows the same read after it has been re-basecalled using standalone Guppy. Now there is Event/Move table under the freshly-added basaecall group (Basecall_1D_001). find_tails() should be called with basecall_group set to "Basecall_1D_001" as shown in the use case # 4 above.

• For DNA data, tailfindr decides whether a read is poly(A) or poly(T) based on finding Nanopore primers/adaptors. If you are using the flipflop model to basecall DNA data, please ensure that the nanopore adaptors are not trimmed off while basecalling. This can be done by turning off enabling_trimming option in the basecalling script. The script below shows you how we have basecalled our reads using the flipflop model

#!/bin/sh
INPUT=/raw/fast5/files/path/
OUTPUT=/output/folder/path/
guppy_basecaller \
    --config dna_r9.4.1_450bps_flipflop.cfg \
    --input $INPUT \
    --save_path $OUTPUT \
    --recursive \
    --fast5_out \
    --hp_correct 1 \
    --disable_pings 1 \
    --enable_trimming 0 

• tailfindr has been tested and validated using the following sequencing kits:

1. SQK-RNA001 and SQK-RNA002 for RNA
2. SQK-LSK108 and SQK-LSK109 for DNA
3. SQK-PCS110 for PCR cDNA

• tailfindr has been tested and validated using the basecallers:

1. Albacore v2.3.1 and v2.3.3.
2. Guppy v2.2.2, v2.3.1, v3.0.1. Guppy v2.2.2, v2.3.1 were tested with flipflop basecalling for DNA, and Guppy v3.0.1 was tested with flipflop basecalling for RNA.

We have written a book chapter on how you can use tailfindr for transcript isofrom-specific poly(A) tail profiling. It has a detailed protocol and analysis pipeline for processing the data through tailfindr .The chapter can be downloaded from here.

If you encounter a clear bug, please file a minimal reproducible example on github. Please do provide us a few reads (around 10) so that we can reproduce the problem at our end, and figure out a solution for you.

Krause, M., Niazi, A. M., Labun, K., Cleuren, Y. N. T., Müller, F. S., & Valen, E. (2019). tailfindr: alignment-free poly(A) length measurement for Oxford Nanopore RNA and DNA sequencing. Rna, 25(10), 1229–1241. doi: 10.1261/rna.071332.119

And of course:

GPL-3: https://www.gnu.org/licenses/gpl-3.0.en.html