guigolab/lyric

Long RNA-seq analysis workflow

LyRic is a versatile automated transcriptome annotation and analysis workflow written in the Snakemake language. Its core functionality is the production of:

1. a set of high-quality RNA Transcript Models (TMs) mapped onto a genome sequence, based on Long-Read (LR) RNA sequencing data.
2. various summary statistics plots and analysis results that describe the input and output data in details
3. an interactive HTML table reporting statistics for each input sample, enabling easy and intuitive sample-to-sample comparison
4. a UCSC Track Hub to display output TMs, as well as various other tracks produced by LyRic.

(Note that features 2, 3 and 4 can be easily switched on and off).

LyRic is platform-agnostic, i.e. it can deal with FASTQ data coming from both the ONT and PacBio platforms.

Full LyRic documentation is here.

lyric's Issues

keep track of unmapped reads

Currently, the "unmapped reads" coming from Minimap2 are discarded by LyRic. Is it possible to create a bam file for "unmapped reads" to tack them?

[Feature request] Configurable genome assembly name for UCSC track hubs

Currently the genome assembly variable for the track hub config is taken from the {genome} wildcard, which is linked to the basename of the fasta file in the local filesystem. The two should be decoupled

Rule makeHtmlSummaryDashboard fails complaining about missing columns

This is the error message:

Error: Can't subset columns that don't exist.✖ Column `intron.x` doesn't exist.Backtrace:     █  1. ├─dplyr::relocate(ntCoverageStats, intron.x, .after = exonOfPseudo.y)  2. ├─dplyr:::relocate.data.frame(ntCoverageStats, intron.x, .after = exonOfPseudo.y)  3. │ ├─base::unname(tidyselect::eval_select(expr(c(...)), .data))  4. │ └─tidyselect::eval_select(expr(c(...)), .data)  5. │   └─tidyselect:::eval_select_impl(...)  6. │     ├─tidyselect:::with_subscript_errors(...)  7. │     │ ├─base::tryCatch(...)  8. │     │ │ └─base:::tryCatchList(expr, classes, parentenv, handlers)  9. │     │ │   └─base:::tryCatchOne(expr, names, parentenv, handlers[[1L]]) 10. │     │ │     └─base:::doTryCatch(return(expr), name, parentenv, handler)
 11. │     │ └─tidyselect:::instrument_base_errors(expr)
 12. │     │   └─base::withCallingHandlers(...)
 13. │     └─tidyselect:::vars_select_eval(...)
 14
Execution halted
Error: Can't subset columns that don't exist.
✖ Column `intron.x` doesn't exist.
Backtrace:
     █
  1. ├─dplyr::relocate(ntCoverageStats, intron.x, .after = exonOfPseudo.y)
  2. ├─dplyr:::relocate.data.frame(ntCoverageStats, intron.x, .after = exonOfPseudo.y)
  3. │ ├─base::unname(tidyselect::eval_select(expr(c(...)), .data))
  4. │ └─tidyselect::eval_select(expr(c(...)), .data)
  5. │   └─tidyselect:::eval_select_impl(...)
  6. │     ├─tidyselect:::with_subscript_errors(...)
  7. │     │ ├─base::tryCatch(...)
  8. │     │ │ └─base:::tryCatchList(expr, classes, parentenv, handlers)
  9. │     │ │   └─base:::tryCatchOne(expr, names, parentenv, handlers[[1L]])
 10. │     │ │     └─base:::doTryCatch(return(expr), name, parentenv, handler)
 11. │     │ └─tidyselect:::instrument_base_errors(expr)
 12. │     │   └─base::withCallingHandlers(...)
 13. │     └─tidyselect:::vars_select_eval(...)
 14
Execution halted
[Thu Jul 28 18:08:30 2022]
Error in rule makeHtmlSummaryDashboard

Update shell prefix command in master.smk

I think the shell.prefix definition in master.smk would need to be updated:

• loading .bashrc is not needed most of the times
• deactivating the active conda environment can give problems (e.g. missing binaries)

Something like:

shell.prefix('set -euo pipefail; export PATH="$PWD/LyRic/utils/:$PATH";')

should probably be fine.

Usee POSIX atomic file install instead of copying from TMPDIR

LyRic would be faster and more robust if it used POSIX atomic file creation instead of writing to TMPDIR and then
copying in place.

TMPDIR is usually on a separate file system, so

mv {TMPDIR}/$uuidTmpOut {output.bam}

Would usually be a copy. If there is an error, such as out of disk space, an incomplete output file
out exist.

If the file is created on the same file system as the output file, so the mv is an atomic rename.
In bash I would do:

tmpout=${outbam}.tmp
do_something  > ${tmpout}
mv -f ${tmpout} ${outbam}

Assuming bash is run with -e, the completed file doesn't exist under
the final name until it is successfully completed, and there is no copy overhead.

lyric can't be checkout on a MacOS

the default MacOS file system is case-insensitive, and lyric has two files that vary only by case:

utils/HashToGff.pm
utils/hashToGff.pm

while running lyric on MacOS is not that useful, it is useful to have a local copy of the source when debugging.