NEWS
dada2 1.16.0
SIGNIFICANT USER-VISIBLE CHANGES
- The 'dada-class$clustering' data.frame includes a new column '$birth_from' which encodes the cluster (ASV) from which the associated cluster (ASV) was divided (or "born").
- Short sequences are now given 'NA' assignments at all levels, rather than stopping the 'assignTaxonomy' function entirely.
BUG FIXES
- All uses of the 'class(x) == "y"' idiom have been replaced by 'is(x, "y")' to conform to current R best practices and Bioconductor build requirements.
- A rare error in how the 'dada-class$clustering$birth_pval' was calculated has been fixed.
- Read lengths are now visualized correctly by 'plotQualityProfile(..., aggregate=TRUE)'.
dada2 1.14.1
BUG FIXES
- A segfault bug in 'assignTaxonomy' has been fixed. This segfault occurred when multithreading was enabled and multiple exact matches to the query sequence existed in the reference database being queried against.
dada2 1.14.0
NEW FEATURES
- Directories containing fastq files (possibly compressed) can now be provided to core dada2 functions instead of a character vector of the fastq filenames. This functionality is supported by 'filterAndTrim', 'learnErrors', 'dada', 'mergePairs' and 'derepFastq'. Note, this feature requires fastqs in the provided directory to have standard file extensions: '.fastq', '.fastq.gz' or '.fastq.bz2'.
- The new 'DETECT_SINGLETONS' option removes the “conditional” in the calculation of probabilties used in the core dada algorithm, which effectively discounts the first read of any novel sequence. In practice, setting 'DETECT_SINGLETONS = TRUE' allows singletons to be detected (of course) and also increases sensitivity to other low abundance sequences slightly, i.e. those present in just 2/3/4 reads. Note, we do not generally recommend this option as it will result in a large increase in false positives in typical datasets. Instead we recommend 'pool = "pseudo"' or 'pool=TRUE' for typical datasets to increase sensitivity to rare sequences with less impact on specificity.
SIGNIFICANT USER-VISIBLE CHANGES
- The 'removePrimers' function has been improved in several ways. Indels are now allowed when matching the primers with the 'allow.indels=TRUE' flag, which can increase primer matching but at a roughly 4x cost in speed. Multiple files are now properly handled, and a previous bug in handling the absence of a reverse primer sequence has been rectified. Note, 'removePrimers' is still only recommended for PacBio or other long-read technologies for speed reasons. For deeper short-read data (e.g. Illumina) we recommend external solutions such as cutadapt or trimmomatic.
- Sequence lengths up to 9999 nucleotides are now supported throughout the dada2 package.
- The new 'tryRC' option in the 'mergeSequenceTables' function will collapse together sequences that are identical up to reverse-complementation. This is most useful for combining datasets from the same gene region, but that may have been sequenced in different orientations.
BUG FIXES
- 'collapseNoMismatch' now properly collapses sequence together that substantially vary in length.
- 'getSequences' now coerces sequences to upper case, as expected by other dada2 functions.
dada2 1.12.0
NEW FEATURES
- The new 'seqComplexity' function quantifies the complexity of sequences in terms of the Shannon richness of their kmers. 'plotComplexity' interrogates the distribution of sequences complexities in fastq files, and the 'rm.lowcomplex' argument in the 'filterAndTrim' function allows filtering of low complexity sequences.
- The new 'removePrimers' function removes forward and reverse primers from sequencing reads, and can orient reads based on the location of the forward primer. Currently we recommend removePrimers for use with PacBio CCS data, but external solutions remain recommended for Illumina data.
SIGNIFICANT USER-VISIBLE CHANGES
- The 'dada' function can now accept fastq filenames rather than requiring files be dereplicated and stored into memory first. This allows memory requirements to remain flat when processing large numbers of samples.
- Pseudo-pooling, an algorithmic approximation to sample inference from pooled samples, now has memory requirements that remain flat with sample number when invoked using filenames, e.g. 'dada(fastqFiles, err=err, pool="pseudo")'. We now recommend pseudo-pooling for those interested in detecting singleton ASVs in their samples.
BUG FIXES
- Pooled sample inference with 'dada(..., pool=TRUE)' no longer fails to output the most abundant ASV in the first sample under certain conditions.
- The data.frame returned by 'mergePairs' is now properly formatted even when only one sample was provided as a list.
dada2 1.9.2
NEW FEATURES
- PacBio CCS reads up to 3 kilobases are now supported. See also PacBioErrfun, the new and recommended error-estimation function for PacBio CCS data.
SIGNIFICANT USER-VISIBLE CHANGES
- primer.fwd has been replaced by orient.fwd in the filterAndTrim function. This option consistently orients mixed-orientation single-end or paired-end reads based on matching the provided sequence fragment to the start or end of each read (or paired read). Intended for use with mixed-orientation reads that included sequenced primers. If primers aren't included in the amplicons, an external re-orientation solution remains preferable.
- trimRight has been added to the filterAndTrim function. This removes the specified number of bases from the end ("right" side) of each read.
BUG FIXES
- collapseNoMismatch now collapses identical sequences as well (previous behavior is togglable).
dada2 1.9.1
BUG FIXES
- mergePairs now gracefully handles cases when zero reads succesfully merge.
- plotQualityProfile now works correclty when given a directory containing fastq files.
dada2 1.7.8
SIGNIFICANT USER-VISIBLE CHANGES
- nbases has replaced the nreads parameter in the learnErrors function. As suggested by the name, this controls the amount of data the machine learning uses by the total number of bases rather than the read count, which is more appropriate given the range of read-lengths in target applications.
- OMEGA_C has been set to 1e-40 by default. This means that error-correction is no longer performed on all reads, but instead just those a post-hoc probability less than OMEGA_C=1e-40. In practice this has a very small impact on final abundances.
BUG FIXES
- The memory usage and speed of assignSpecies on large datasets has significantly improved.
dada2 1.7.7
NEW FEATURES
- Error-correction can now be modified, and turned off, by the OMEGA_C parameter, which controls the threshold at which reads that are inferred to contain errors are corrected (or not) to the sequence from which they are inferred to originate.
SIGNIFICANT USER-VISIBLE CHANGES
- The DADA2 options enabling the quick gapless alignment check, and extremely conservative greediness in the partioning method were turned on by default (GAPLESS=TRUE, GREEDY=TRUE). Some speedup in the core denoising algorithm.
- plotQualityProfile now includes a cumulative description of read length variation.
dada2 1.7.6
NEW FEATURES
- A new and extremely conservative form of greediness in the core denoising algorithm was added, and can be turned on by setting the DADA2 option GREEDY=TRUE. This provides some speedup in the core denoising algorithm.
dada2 1.7.5
NEW FEATURES
- The dada(...) function now accepts a list of "priors", i.e. sequences for which there is prior evidence they might be real. Input sequences that match one of the priors are evaluated against a relaxed threshold of statistical evidence (OMEGA_P instead of OMEGA_A), and can be detected even as singletons.
- The dada(...) function can perform "pseudo-pooling" with dada(..., pool="pseudo"). In pseudo-pooling, the input samples are denoised independently, then a set of sequences that appear in at least MIN_PREVALENCE samples are used as priors for a second and final round of sample inference. MIN_PREVALENCE=2 by default.
dada2 1.7.4
NEW FEATURES
- A new fast screen for optimal gapless alignments in the core denoising algorithm was added, and can be turned on by setting the DADA2 option GAPLESS=TRUE. This provides some speedup in the core denoising algorithm.
dada2 1.7.3
BUG FIXES
- Fixed an overflow bug on sequences 260nts or longer in the SSE=2 code.
dada2 1.7.2
SIGNIFICANT USER-VISIBLE CHANGES
- The DADA2 option enabling explicit 8-bit SSE vectorization in the C code was turned on by default (SSE=2). Some speedup in the core denoising algorithm.
dada2 1.7.1
NEW FEATURES
- seqComplexity calculates the complexity of input sequences, and can be used to identify and filter out low-complexity sequences.
SIGNIFICANT USER-VISIBLE CHANGES
- The default minOverlap parameter of mergePairs was reduced from 20 to 12, and the alignment parameters used during merging were altered to more strongly penalize mismatches and gaps, which improves merging performance in repetitive sequences.
dada2 1.5.6
SIGNIFICANT USER-VISIBLE CHANGES
- The default values of the minFoldParentOverAbundance parameter changed. Minor impact on removeBimeraDenovo(...) output.
- The new MIN_ABUNDANCE option (default 1) specifies the minimum required abundance for a unique sequence to be output as a denoised sequence variant. Higher values can significantly reduce computation time on large samples, or pooled sets of samples.
BUG FIXES
- collapseNoMismatch now properly collapses together sequences that overhand others sequences on both sides (and is faster).
dada2 1.5.4
BUG FIXES
- End-point off-by-one bug in kmer distance calculations fixed. Minor impact on dada(...) output.
- Multithreading bug fixed.
dada2 1.5.3
SIGNIFICANT USER-VISIBLE CHANGES
- A version of the kmer-distance calculation that is explicitly SSE vectorized has been added and can be toggled on with the dada(..., SSE=1) option. This is intended for use when the code has not been natively compiled by a modern compiler, such as is the case for the bioconda binaries.
dada2 1.5.2
SIGNIFICANT USER-VISIBLE CHANGES
- assignSpecies has been reimplemented to be much, much faster (>100x faster with enough query sequences).
- The allowOneOff flag has been set to FALSE by default in the xxxBimeraDenovo functions.
BUG FIXES
- The compress option is now respected by filterAndTrim.
dada2 1.5.1
SIGNIFICANT USER-VISIBLE CHANGES
- The parsing functions to create the training fastas compatible with assignTaxonomy and assignSpecies from the Silva and RDP databases are now included as private functions in the taxonomy.R file. In parallel, updated versions of both RDP (trainset 16) and Silva (v128) training fastas were released on Zenodo.
dada2 1.3.5
SIGNIFICANT USER-VISIBLE CHANGES
- The default chimera removal method is now "consensus" in the removeBimeraDenovo function (method="consensus").
- The filtering functions now remove phiX by default (rm.phix=TRUE).
dada2 1.3.4
NEW FEATURES
- filterAndTrim is the new multithreaded interface for filtering and trimming. It is R-vectorized, in that it accepts vectors of input/output files. Supports both single-read and paired-read filtering.
SIGNIFICANT USER-VISIBLE CHANGES
- Consensus chimera removal now supports multithreading.
- The new trimOverhang option in mergePairs allows the portion of paired reads that overhang the start of the other read to be removed when merging.
- Faster filtering by not enforcing redundant filtering conditions.
dada2 1.3.3
NEW FEATURES
- ITS support. The dada method now handles variable-length amplicons, which allows common ITS sequencing strategies to be processed. Taxonomic assignment against the UNITE database is also now available.
SIGNIFICANT USER-VISIBLE CHANGES
- The xxxFilter methods now have maxLen and minLen filtering parameters.
dada2 1.3.2
NEW FEATURES
- learnErrors is a new method for learning error rates from a targeted number of reads. This wraps pre-existing functionality -- dada(..., selfConsist=TRUE) -- in a much easier to use form.
- assignTaxonomy is multithreaded (multithread=TRUE) and can individually test reverse-complemented sequences and use them to classify if they better match the reference (tryRC=TRUE).
dada2 1.3.1
SIGNIFICANT USER-VISIBLE CHANGES
- The xxxFilter methods now (invisibily) return the number of reads that went in and that passed the filter.
- fastqFilter now has a maxLen parameter, primarily for filtering/trimming 454 data.
BUG FIXES
- trimLeft is now enforced before any filtering is performed.
- Rscript invocations should now work without requiring an initial library(methods) call.
dada2 1.1.7
SIGNIFICANT USER-VISIBLE CHANGES
- The isBimeraDenovoTable method is now available through the convenience interface removeBimeraDenovo(..., tableMethod="consensus").
- The vectorized Needleman-Wunsch algorithm is now available through nwalign(..., vec=TRUE). This aligner is very fast for short sequences, but performs no overflow checking so use on longer sequences (2kb+) with care.
BUG FIXES
- It is now more difficult to mess up the internal structure of dada-class and derep-class objects.
dada2 1.1.6
NEW FEATURES
- isBimeraDenovoTable is a new function for detecting chimeras on multi-sample sequence tables. It first identifies bimeras on a per-sample basis, and then uses a consensus approach to identify bimeric sequences. This improves the specificity of bimera classification for datasets with many samples.
SIGNIFICANT USER-VISIBLE CHANGES
- The filenames is now included in the plotQualityProfile plot.
BUG FIXES
- assignTaxonomy now works for sequences containg Ns.
- Bimera detection now correctly handles non-unique sequences in the input.
- Fixed a memory leak when using the HOMOPOLYMER_GAP_PENALTY option.
- Warnings and errors during error estimation from small samples in the dada(…) function are now handled appropriately.
dada2 1.1.4
NEW FEATURES
- Multithreaded chimera detection. The isBimeraDenovo(...) function is now multithreaded. This behavior is controlled by the multithread argument to the function, and is FALSE by default.
dada2 1.1.3
SIGNIFICANT USER-VISIBLE CHANGES
- The vectorized aligner has been updated. It is now more flexible, handling different length sequences and variable end-gap penalties. It is also faster: On default settings the dada(...) function is now 15-25% faster, and isBimeraDenovo is 2-4x faster.
dada2 1.1.2
NEW FEATURES
- Species-level taxonomic assignment. The new assignSpecies(...) function uses exact matching to assign sequences to the species level. Currently a valid species training fasta is only available for the RDP taxonomic database.
dada2 1.1.1
BUG FIXES
- Fixed the memory leak in isBimeraDenovo and friends.
- Sequence tables are now integer rather than numeric.
dada2 1.1
NEW FEATURES
- Multithreading. The dada(...) function is now multithreaded. This behavior is controlled by the multithread argument to the function, and is FALSE by default.
SIGNIFICANT USER-VISIBLE CHANGES
- The phiX removal functionality in fastqFilter/fastqPairedFilter is now substantially faster.
BUG FIXES
- Certain diagnostic return values of dada(..., pool=TRUE) are now appropriately constructed.
- A 2-5% speedup from switching to C++11 hashed containers.
dada2 1.0
NEW FEATURES
- The dada2 package is now available from Bioconductor (http://bioconductor.org)
dada2 0.99.9
NEW FEATURES
- fastqFilter and fastqPairedFilter can now remove phiX contamination if rm.phix argument set to TRUE
dada2 0.99.8
SIGNIFICANT USER-VISIBLE CHANGES
- Banding in nwalign is now turned off by default
BUG FIXES
- Graceful handling of sequences which are not classified at any level by assignTaxonomy
dada2 0.99.7
SIGNIFICANT USER-VISIBLE CHANGES
- isBimera was rewritten in C, significantly increasing speed
BUG FIXES
- An edge case bimera detection bug was fixed
dada2 0.99.6
SIGNIFICANT USER-VISIBLE CHANGES
- Function documentation was reviewed and revised throughout the package
dada2 0.99.5
NEW FEATURES
- plotQualityProfile displays a visual summary of the quality scores over sequences in a fastq file
- removeBimeraDenovo conveniently identifies and removes chimeras from the input unique sequences
SIGNIFICANT USER-VISIBLE CHANGES
- The dada2 package is now part of the devel branch of Bioconductor (http://bioconductor.org)
BUG FIXES
- assignTaxonomy now handles varying levels of taxonomic classification in the training data
dada2 0.10.7
NEW FEATURES
- dada2 now supports 454 pyrosequencing. When calling the dada function on 454 data, we recommend
using the parameters USE_QUALS = FALSE, HOMOPOLYMER_GAP_PENALTY = -1, BAND_SIZE = 32
SIGNIFICANT USER-VISIBLE CHANGES
- mergePairs is now "vectorized" over lists of input dada-class and derep-class objects
- Added a homo_gap argument to nwalign which sets the homopolymer gap penalty in the N-W alignment
BUG FIXES
- assignTaxonomy now uses less memory