All notable changes to this project will be documented in this file.
The format is based on Keep a Changelog.
- FamDB is now a dependency of RepeatModeler and has been updated to support Dfam 4.0
- Configure now logs without screen clearing making it easier to debug problems.
- RepeatClassifier is now responsible for querying and caching reference libraries from FamDB.
- Bug that caused RepeatModeler to fail when used with the older RECON 1.08 release. Github Issue #312
- Support for makeblastdb partitioning of large genomes (e.g.
*.00.nsq). - Long format support in 2bit files for large genomes.
- Catch for older versions of RepeatScout.
- Support for RECON 1.10.
- RepeatModeler/Refiner was not processing consensus extensions properly in 2.0.7 and in some cases was concatenating forward and reverse strand sequences together, adding noise to seed alignments and consensus calls.
- RepeatAfterMe auto-extension step to Refiner, applied to both RepeatScout and RECON generated family models.
- RepeatClassifier now checks for a complete Dfam FamDB database.
- New redundancy filter applied to RepeatScout families.
- Lowered masking score threshold to reduce rediscovery.
- Bug with the recoverDir round detection logic.
- MultAln: new API for calculating Kimura divergence.
- MultAln: new API for slicing operations on MSA.
- MultAln: switch to Smitten V2 for default STK output.
- MultAln: improvements to the Linup reporting format.
- Refiner reorganization for future extension algorithm.
- Updated to RepeatScout 1.0.7, which reports extended seed coordinates, reducing complexity when searching for exemplars using the consensus sequence.
- Reduced the minimum family instances for RepeatScout families.
- Bug that caused RepeatClassifier to run with only one thread in the last release.
- Bug that caused some RepeatScout-derived exemplars to be reported on the wrong strand.
- Long sequence identifiers from LTR_retriever that altered names longer than 13 characters are now isolated correctly.
- Fixes for LTR_retriever changes in TRF parameter.
-longoption tofaToTwoBitto support larger genomes.
- Bug that caused failures in "absolute" reproducibility. Prior to this release,
use of
-srandwould not guarantee thatconsensi.fa.classifiedandfamilies-classified.stkwere identical in sequence and sequence order across runs. Secondary sorts were added to guarantee a fixed sort order among equally scoring results. Note: This change applies only to results generated with this version and future releases.
- New
threadsoption replacing thepa(parallel batches) option, mapping directly to the maximum number of threads the program will attempt to use. - Support for RepeatMasker 4.1.4 and RMBlast 2.13.0 parallel query feature.
- Larger default sample sizes enabled by speed improvements; original sampling
strategy selectable with the new
quickoption.
- Improved gathering of RepeatScout exemplars for building seed alignments.
- Parallelized and improved masking between rounds for faster runs and fewer rediscoveries.
- ABBlast is no longer supported.
- Several visual artifacts in the
viewMSA.plHTML visualization.
- Program now generates a logfile in the working directory named
<seqfile>-rmod.log, containing the random seed number and high-level run stats for use when reporting problems.
- Problem with orientation/coordinates in Stockholm output format that affected a subset of sequences.
- Bug affecting the trim functions of the Linup tool.
- First release of a set of manual curation tools for use with de-novo generated TE libraries.
generateSeedAlignments.plto generate Dfam-compatible seed alignments given a consensus-based TE library and RepeatMasker output.
- Bug in N50 calculation.
- Ruzzo-Tompa maximal scoring subsequences implementation.
- Various small bugfixes.
- Version of each dependency is now printed in the log.
- Workaround for a bug introduced in NCBI Blast 2.10.0 with version 5 databases.
- RepeatScout in rare cases generates models for very long-period satellites, causing Refiner to produce large numbers of off-diagonal alignments; these cases are now filtered out.
- Container releases (Docker and Singularity).
- Code to remove excess temporary files from the
RM_directory after they are no longer needed.
- Seed alignments and consensus sequences previously contained stretches of N's due to TRF masking; original genomic sequences are now used instead, preserving low-complexity regions in the final consensus.
- Batching system refactored to improve handling of short input sequences common in early assemblies or low-coverage genome projects.
- TRF interface refactored to eliminate a major performance bottleneck in the simple/tandem repeat masking step.
- Stockholm sequence coordinates did not consistently show strandedness via coordinate reversal.
- Temporary file generation in
BuildDatabasedid not use a unique filename, risking race conditions when twoBuildDatabaseprocesses ran in the same directory simultaneously. - Bug that caused some TE families to be labeled with the RepeatMasker internal-use label "buffer".
- TRF masking of input sequences caused some low-complexity sequences to incorrectly match TE library sequences.
- Conflict resolution often let a lower-scoring match define the class for a family.
- Avoid bailing if Refiner could not produce a consensus from the family sequence set.
- Bug that caused RepeatModeler to exit if no new repeats were detected after round 2. This may occur with small sample sizes or repeat-poor sequence sets and does not necessarily mean repeats will be absent in later rounds.
- Random number generator seed is now printed at the start of a run, enabling
exact reproduction of genome samples via the
-srand ####flag in future runs. - Final output files are now placed in the same directory as the input database.
- Additional output file containing the seed alignment for each discovered family,
stored in a Dfam-compatible Stockholm file. New output files are named
<database_name>-families.faand<database_name>-families.stk. - Support for Dfam_consensus built into this release. Utilities
dfamConsensusTool.plandrenameIds.plare available in theutil/directory.
- New
-recoverDir <dir>option allowing RepeatModeler to restart where it left off after a failure. - BLAST searches can now be run in parallel on shared-memory multiprocessor
machines via the new
-pa #option.
- Support for RMBlastn version 2.2.27.
TRFMaskupdated to work with the new RepeatMasker 4.0 release.- New
viewMSA.plutility.