List of changes from version 3.3.2 to 3.3.3 (until Sep 13th, 2019)
- docker container
- new script tfix_in_frame_stop_codon_genes.py that replaces genes where spliced in-frame stop codons were predicted
- new scripts: compare_masking.pl, merge_masking.pl
- new program bamToWig.py as alternarive to bam2wig C++ binary
- fix warnings on new GCC compiler (8.3)
- introduced unit tests of C++ code (make test)
- bugfixes (lex.cc, autoAug.pl)
- DIAMOND as alternative to BLAST in aa2nonred.pl
List of changes from version 3.3.1 to 3.3.2 (until Oct 5th, 2018)
- bugfixes in comparative augustus, utrrnaseq
- new species Chiloscyllium punctatum (bamboo shark), Scyliorhinus torazame (cat shark), Rhincodon typus (whale shark)
- updated comparative augustus (CGP) tutorial
List of changes from version 3.3 to 3.3.1 (until May 8th, 2018)
- new species pisaster (Pisaster ochraceus, ochre starfish)
- bugfixes of sampling error in intron model and in check of
frame-compatibility of CDSpart hints without frame
- new auxiliary program utrrnaseq to find training UTRs
- new script setStopCodonFreqs.pl
List of changes from version 3.2.3 to 3.3 (until July 11th, 2017)
- new program ESPOCA to estimate selective pressure on codon alignments
- gene finding on ancestral genomes is enabled
- new default parameters for comparative gene prediction (CGP)
- clade parameters training for CGP
- compatibility to Ian Fiddes' Comparative Annotation Toolkit (CAT)
- new scripts eval_dualdecomp.pl,
- more tolerant tree parsing
- bugfixes in augustus, joingenes, load2sqlitedb, transMap2hints.pl, splitMfasta.pl, intron2exex.pl,
aln2wig
- new functionality in homGeneMapping, joingenes
- new species ciona, mnemiopsis_leidyi, nematostella_vectensis, strongylocentrotus_purpuratus
List of changes from version 3.2.2 to 3.2.3 (until December 6th, 2016)
- bugfixes in joingenes, load2sqlitedb, transMap2hints.pl
- added script gtf2bed.pl
- CGP parameters for clade of 8 primates (cgp/log_reg_parameters_primates.cfg)
- added option --clean to load2sqlitedb
- now filter out truncated coding genes with CDS length <= 3
- Sascha Steinbiss: remove dependency on PATH_MAX
List of changes from version 3.2.1 to 3.2.2 (until March 30th, 2016)
- added extrinsic.M.RM.PB.cfg for PacBio CCS hints with GMAP
(see http://bioinf.uni-greifswald.de/bioinf/wiki/pmwiki.php?n=Augustus.PacBioGMAP )
- standardization of Makefile
- improvements in auxiliary program joingenes
- added missing program load2sqlitedb
- comparative gene predictions (CGP):
- hints with source=M (e.g. existing annotations) are fixed
- new standard parameters for exon_gain, exon_loss and others
- bugfixes, in particular a double free or corruption bug in PPX that occured with BUSCO
- new species Volvox carteri (volvox)
List of changes from version 3.2 to 3.2.1 (until October 23rd, 2015)
- comparative gene prediction (CGP): make the parameters from the GCB proceedings paper match again
- new CGP default parameters (tested on 5 clades)
- CGP version penalizes very long introns equivalently to a geometric distribution
- code now under artistic license only
List of changes from version 3.1 to 3.2 (until September 25th, 2015)
- comparative gene prediction (CGP option) has sensible default parameters of the logistic regression model
- new tool hal2maf_split.pl for splitting an alignment for parallel comparative prediction
- bugfixes to augustus (compiler warnings on MaxOS)
- improvements to prediction of noncoding genes
- homGeneMapping can determine homologous sets of transcripts
- bugfixes in joingenes and bam2hints (rare segmentation fault)
- made load2sqlitedb more robust, new option to skip index building when loading many species
- new tutorial that includes iterative training
- new option to shift coordinates in hint input when sequence window is cut out
- adapted cegma2gff.pl to cegma version 2.5
List of changes from version 3.0.3 to 3.1 (until April 29th, 2015)
- prediction of noncoding genes (ncRNA) with option --nc=on, requires input of hints, e.g. from RNA-Seq,
predicts coding genes (with UTRs) and noncoding genes simultaneously
- speed improvement with --UTR=1 (by 83% on human without sampling, by 18% on fly with sampling)
- softmasking support: --softmasking=1 is a little more accurate and much faster on human
- small overall accuracy improvement
- prevent the prediction of spliced in-frame stop codons (for SHORT or HINTED introns only)
- new species: rice, Danio rerio (zebrafish), s_aureus (Staphylococcus aureus),
camponotus_floridanus (ant) contributed by Shishir K Gupta,
sulfolobus_solfataricus (archaeon), Parasteatoda tepidariorum (house spider), Apis dorsata (from Francisco Camara Ferreira)
- update to parameters of Thermoanaerobacter tencongensis, E.coli, Yarrowia lipolytica, schistosoma2 (Schistosoma mansoni)
- remove dependency on libboost-filesystems, boost is now completely optional in single-species mode
- new auxiliary program 'homGeneMapping' for identification of homologous genes in a set of gene predictions
- new auxiliary program 'joingenes' for combining gene sets
- bugfixes in load2db, length distribution skew of bacteria, comparative mode (findGeneRanges,
Alignment::mergeable, memory leak), some ass hints were previously not used although possible
- decreased minintronlen default of bam2hints to 32 as the old threshold tended to be too large
- computing of omega = dN/dS more efficiently
- hints can be loaded to the sqlite database
- update of bam2wig to latest samtools
List of changes from version 3.0.2 to 3.0.3 (until August 28th, 2014)
- made dependency on boost libraries an option
- softmasking support (--softmasking=1 improves speed by almost a factor of 2 and exon level sensitivity by 0.01 compared to hard masking on human)
- comparative gene finding with UTRs
- new sqlite database option for genome input in comparative mode besides flat files and MySQL
- new species: zebrafish (Danio rerio), camponotus_floridanus (an ant) contributed by Shishir K Gupta, s_aureus (Staphylococcus aureus)
- updated parameters: thermoanaerobacter_tengcongensis, E_coli_K12
- bugfix in bacterial model: length probability of very long genes was wrong
- other bugfixes in comparative mode
- removal of mysql++ code from package: this is available as Debian/Ubuntu package
List of changes from version 3.0 to 3.0.2 (until February 22nd, 2014)
- inclusion of scanner and parser code for Newick tree format (GPL license), easier installation of
comparative augustus functionality
- therefore release under more restrictive GPL license (the plan is to replace this later
and switch back to the more permissive and non-viral artistic open source license)
- drop requirement to set environment variable AUGUSTUS_CONFIG_PATH
- bugfixes (etraining: "not Genbank format", gtf2gff.pl, compilation warning)
List of changes from version 2.7 to 3.0 (until January 8th, 2014)
- comparative gene prediction (cgp) mode to predict genes in several aligned genomes simultaneously (see README-cgp.txt)
- load2db program to load genomes and hint GFF files into a database for random access retrieval (e.g. with new program getSeq)
- new parameter --temperature for heating the distribution and obtaining more gene structures during sampling
- bacterial genes may now also be nested
- new script weedMaf.pl that throws out gap-only columns from MSAs
- added prokaryotic template parameters for training new prokaryotic species (on the basis of E_coli_K12 parameters)
- new species coyote_tobacco (Nicotiana attenuata), thermoanaerobacter_tengcongensis, wheat
- ATG is not anymore the only start codon
- filterMaf.pl takes an alignment in MAF format and extracts alignment blocks given a subset of species or a genomic interval
- bugfixes (in particular segfault on MAC, small memory leaks)
List of changes from version 2.6.1 to 2.7 (until March 4th, 2013)
- support for prokaryotic genomes (overlapping genes, arbitrary sets of start codons)
--genemodel=bacterium still preliminary
- new species chicken, xenoturbella, heliconius_melpomene1, cacao (Don Gilbert)
- improved species tribolium2012
- speedup of bam2hints, filterBam (by a very large factor in cases of large headers)
- no in-frame codon-split stop codons anymore with option --mea=1
- new script for converting CEGMA gff output format to a gff format
- database access as alternative to flat file input for comparative gene finding
- bugfixes (--noprediction, bam2hints, .gb input with Windows line breaks, tetrahymena,
optimize_augustus)
List of changes from version 2.6 to 2.6.1 (until June 9th, 2012)
- make the sources compatible with older compilers that do not understand the flag -std=c++0x
- script samMap.pl
- filterBam bugfixes
- customized transition matrix for species chlamy2011
List of changes from version 2.5.5 to 2.6 (until April 12th, 2012)
- new species bombus_terrestris, rhodnius, Conidiobolus_coronatus
- new auxiliary programs filterBam, bam2hints
- update to species tomato
- option --mea=1 to maximize expected accuracy instead of using Viterbi algorithm
- new option --contentmodels=off that neutralizes all models except signal models
(to analyse SNP influence on gene structure)
- new parameter --min_intron_len for species with shorter introns than the default 39bp
- added scripts wigchoose.pl, filterShrimp.pl
- updated RNA-Seq documentation readme.rnaseq.html
- bugfixes (in augustus, autoAugPred.pl, maskNregions.pl, optimize_augustus.pl, filterPSL.pl,
intron2exex.pl
List of changes from version 2.5 to 2.5.5 (until March 2nd, 2011)
- added a tutorial
- bugfixes, in particular one that caused a crash on Mac OS
- new species verticillium, honeybee, butterfly
- option --noprediction=1 to score input gene structures
- adjust autoAug.pl to new PASA version
List of changes from version 2.4 to 2.5 (until Nov 13th, 2010)
- first release of protein profile extension (PPX)
- scripts added for PPX:
msa2prfl.pl: converts MSAs in FASTA/CLUSTAL format to protein profiles to use with AUGUSTUS
block2prfl.pl: converts BLOCKS database flat file to protein profile
del_from_prfl.pl: manually delete blocks from profile
- executables added for PPX:
prepareAlign: eliminates gap-rich sequences from FASTA MSA
fastBlockSearch: performs a preliminary profile search to determine regions for gene prediction
- PPX-specific new parameters are:
/ProteinModel/allow_truncated: allows profile hits in right-truncated genes
/ProteinModel/block[part]_threshold_{sens,spec}: set model specificity/sensitivity
/ProteinModel/weight: set weighting of protein model vs. ab-initio model
- PPX is introduced and can be used with parameter --proteinprofile
- new species pneumocystis (fungus), user contributed by Marco Pagni
List of changes from version 2.3.1 to 2.4 (until July 27, 2010)
- improved human parameter set: roughly 13 percent points better on CDS exon level
- improved fly parameter set
- added new species 'lamprey', contributed by Falk Hildebrand and Shigehiro Kuraku
- added new species Leishmania tarentolae, Trichinella spiralis
- added maize5 parameter set, better than 'maize' and with UTR model
- the GC content parameter class is now chosen locally, instead of using a constant
GC constant for each sequence or sequence chunk, performance improvement in human
- bugfix in aln2wig that caused a segmentation fault
- bugfix in etraining when using alternative translation tables
- introduce local splice site malus, that is effective when neighboring region has exonic evidence
- bugfix in etraining that concerns the weighting of training examples
the fix increases e.g. in C.elegans accuracy by 1.5 percent points on the exon level
and 4.2 percent points on the gene level
- added new script gtf2gff to convert gtf output (e.g. from Augustus) to gff3
or to switch between the two UTR representations (explicit UTR vs exon)
- added script intron2exex.pl for iterative RNA-Seq alignments
- option --pairbed to filterPSL.pl for creating genicpart hints from read pairing information
- allow to tie intergenic content model to intron content model with tieIgenicIntron in training
- bugfix: intron/igenic content model was partially trained from UTR regions
- fixed a bug that occurred when a sequence contained nothing but N's
- fixed bug in intronmodel training that occurs when UTRs are turned on
- fixed pasa to polyA hints bugs in autoAug.pl
List of changes from version 2.2 to version 2.3 (until November 1st, 2009):
- added capability and documentation for incorporating RNA-Seq
- added aln2wig program which summarizes psl and shrimp format to wig
- speed up script makeUtrTrainingSet.pl
- hints can now have a mult=n feature in the last column to specify multiplicity=copy number
- added local malus for punishing unevenly supported exon candidates
- gff2gbSmallDNA.pl now allows for gff3 input
- getAnnoFasta.pl now also prints coding sequence and mRNA derived from genome sequence
- script polyA2hints.pl for hints from polyA position file
- updated blat2hints: for intron hints from RNA-Seq and for transcript termini
- parameter /IntronModel/dss_consensus_gc_prob now allows gc-ag introns also in absense of hints
- autoAug pipeline now includes polyA hints from PASA
- parameter allow_hinted_splicesites allows other non-standard splice sites, such as ac-at
- bugfixes (in particular in response to MAC users compilation problems)
List of changes from version 2.1 to version 2.2 (until June 26th, 2009):
- release of the autoAug*.pl scripts for automatic training and prediction for new genomes
- new genemodel 'intronless' for prokaryotes and eukaryotes withouth introns
List of changes from version 2.0.3 to version 2.1 (until March 11th, 2009):
- updated species: Chlamydomonas reinhardtii
- new species: Acyrthosiphon pisum (pea aphid)
- packaged more scripts that are useful for training
- new hint type genicpart = nonirpart
- introduce fuzzyness in dss and ass hints
- reduced memory usage
- allow other poly(A) signal consensus than aataaa
- fixed some format violations when outputting gff3
- allow general alternative codon to amino acid translation tables (parameter translation_table)
List of changes from version 2.0.2 to version 2.0.3 (until March 13th, 2008):
- new species: Chlamydomonas reinhardtii (green algae)
- new species: Nasonia vitripennis (wasp)
- new species: Solanum lycopersicum (tomato)
- new species: Amphimedon queenslandica (sponge)
- in fasta headers (">") the name is considered only to go up to the first white space
List of changes from version 2.0.1 to version 2.0.2 (until May 30th, 2007):
- a gene can now be truncated in the middle of an exon or an intron.
This is helpful when many genes are only partially sequenced, e.g. when searching genes/exons
in the raw DNA reads.
- Augustus retrained for Arabidopsis thaliana, Galdieria sulphuraria and Caenorhabditis elegans.
- fixed bug that no spliced 5'UTR was predicted on the reverse strand except at sequence boundaries
- optimize_augustus.pl now also allows to automatically optimize the transition matrix
- fixed bug. before some hints were ignored when the contig was shorter than 2000bp
- option --uniqueGeneId=true adds the sequence name to the gene name
List of changes from version 2.0 to version 2.0.1 (until January 30th, 2007):
- genemodel=complete works now with UTR
- added several scripts for generating hints (from blat, retrogenes, transmap, exoniphy) and
for creating cluster jobs
- fixed bug with intronpart hints upvaluing a few non-intron bases in UTR
- can now combine --genemodel=complete with UTR predictions
- fixed crash that occurred when a genbank file had annotation with negative coordinates.
- fixed problem with toxoplasma UTR transition matrix (added the 4 extra UTR intron states)
- fixed problem with galdieria UTR model (TTSMOTIF was missing)
List of changes from version 1.8.2 to version 2.0 (until January 8th, 2007):
- predict 5'UTR and 3'UTR including introns (available for human, galdieria, toxoplasma,
caenorhabditis)
- evidence-based alternative transcripts
- new types of hints: tss, tts, CDS, CDSpart, UTR, UTRpart, irpart, nonexonpart
- hints can be summarized to groups
- hints can have a priority
- long input sequence is split taking hints into account
- 32 new fungi species
- output in gff3 format possible
- speed improvement by 12% (comparison operator)
- a subwindow for predictions can be specified
- subdirectory structure of config directory
- new species: toxoplasma gondii
List of changes from version 1.8 to version 1.8.2 (until March 10th, 2006):
- new species: Zea mays (maize)
- Species parameters contributed by Jason Stajich: Caenorhabditis
elegans, Saccharomyces cerevisiae, Ustilago maydis, Phanerochaete
chrysosporium, Neurospora crassa, Histoplasma capsulatum, Fusarium
graminearium, Cryptococcus neoformans, Aspergillus nidulans
List of changes from version 1.7 to version 1.8 (until January 18th, 2006):
- new species: Schistosoma mansoni, Tetrahymena thermophila, Galdieria sulphuraria
- enabled non-standard set of stop codons (in Tetrahymena thermophila TAG and TAA are coding)
- frequencies of the 3 stop codons configurable
- option: AUGUSTUS_CONFIG_PATH can be set on the command line
- some filtering of the Viterbi prediction with respect to posterior probabilities
List of changes from version 1.6 to version 1.7 (until October 5th, 2005):
- automatized training for new species
- alternative transcripts
- posterior probability of exons, introns, transcripts and genes is reported
- fixed bug that predicted splice sites just outside the sequence boundaries
- added the option of outputing the predicted coding sequence
List of changes from version 1.5 to version 1.6 (until August 3rd, 2005):
- The source code is now freely available (Artistic License).
- New species: Tribolium castaneum, Aedes aegypti, Coprinus cinereus
- Bugfix: Wrong interpretation of reading frame for reverse strand exon hints.
- Training of hints possible with nested genes.
- Renaming StateType 'single' to 'singleG' to avoid compilation conflicts on some machines
- Bugfix: Exon hints and exonpart hints including the stop codon were missed for reverse terminal or reverse single exons.
- Changed output format to gtf. Includes a (redundant) transcript_id now.
- Check consistency of hints with the sequence and output warnings for inconsistencies.
- Bugfix: Rare bug in aSSProb that caused Viterbi to get stuck.
- Bugfix: Check sequence bounds, when probing for a splice site. Bug had caused a segmentation fault.
List of changes from version 1.4 to version 1.5 (until 12/04):
- New species: Brugia malayi.
- Training can now take sequences with multiple genes on both strands.
- The use of hints has been documented.
List of changes from version 1.3 to version 1.4 (after 06/04):
- New species: Arabidopsis thaliana.
- More standard, user friendly program call. No need to specify a configuration file anymore.
- Fixed bug for the emission probabilities of exonpart pitchfork hints in exons which are supported
by exonpart hints. Now AUGUSTUS has an incentive to predict an exon which is as small as possible even
when an exonpart hint supports it.
- Fixed rare bug with exon state at end of DNA.
- Optimized meta parameters for human.
List of changes from version 1.2 to version 1.3:
- avoid predicting exons with in-frame stop codons at the splice sites
- accuracy of extrinsic information can now be checked on multi-gene sequences
List of changes from version 1.0 to version 1.2:
- Added reference information to the text output.
- Bugfix: The last line of an input file which contains no newline at the end was
read twice.