************************************************** * 2.10 SERIES NEWS * ************************************************** BASIC CONTAINERS o Added a set of "coerce" methods for turning an arbitrary XStringSet object into a BStringSet, DNAStringSet, RNAStringSet or AAStringSet instance (via the as() function). o Added an "append" method for XStringSet objects. An important use case for this is to put together a set of short reads and their reverse complements in a single DNAStringSet object and then to turn this object into a single PDict object (dual PDict object). Then this dual PDict object can be used to walk each reference sequence only once (instead of twice) in order to get the hits in both strands (+ and -). o Removed the XStringList class and family. o Moved the IRanges, UnlockedIRanges, LockedIRanges, NormalIRanges, MaskCollection, Views, and XInteger classes and their methods to the new IRanges package. C-LEVEL FACILITIES UTILITIES o Added the codons() and translate() generic functions with methods for DNAString, RNAString, DNAStringSet, RNAStringSet, MaskedDNAString and MaskedRNAString objects. o Added the hasOnlyBaseLetters() and uniqueLetters() generic functions and methods. o Added fasta.info() for fast extraction of the descriptions and lengths of the sequences stored in a FASTA file. Also renamed the 'strip.desc' argument of readFASTA() -> 'strip.descs'. o Renamed replaceLetterAtLoc() -> replaceLetterAt() and renamed its 'loc' argument -> 'at'. Deprecated replaceLetterAtLoc(). o Added predefined 'RNA_GENETIC_CODE' object. o Moved the utility functions for importing a mask (read.agpMask(), read.gapMask(), read.liftMask(), read.rmMask() and read.trfMask() functions) to the new IRanges package. o Moved the generic functions for width(), shift(), restrict(), narrow(), reduce(), gaps(), reverse(), coverage(), subject(), views(), trim(), and subviews() to new IRanges package. STRING MATCHING o Added the vcountPDict() generic functions with a method for XStringSet objects. It is the vectorized version of countPDict() i.e. the subject must be an XStringSet object. o Added support for indels to matchPattern(), countPattern() and vcountPattern() (vmatchPattern() will follow as soon as MIndex objects support variable-width matches). o Added the vmatchPattern() and vcountPattern() generic functions with methods for XStringSet objects. They are the vectorized versions of matchPattern()/countPattern() i.e. the subject must be an XStringSet object (support for XStringViews objects will follow soon). o Added matchPWM() and countPWM() methods for XStringViews and MaskedDNAString objects. o Addition of the 'dups0' slot to the ByPos_MIndex class: this allows a more compact representation in memory of a ByPos_MIndex object that holds the hits of a set of patterns that has a lot of duplicates. The benefit is really noticeable when the patterns that are highly represented in the original dictionary have a lot of hits which seems to be typically the case when matching Solexa data against their reference genome. In this case, using the new 'dups0' slot can make the ByPos_MIndex object about 3 times smaller. Take advantage of this new 'dups0' slot to improve the way duplicated patterns are handle thru the "PDict -> matchPDict() -> MIndex" pipe. The new strategy is to "remove them as early as possible and put them back as late as possible". This leads to a gain in speed and also less memory is needed to store the hits in the temporary buffer. o Added the "whichPDict" generic function with a method for XString objects. o Major rework of the PDict class, subclasses and the PDict() constructor: - Merged the CWdna_PDict and TBdna_PDict classes into the TB_PDict class (subclass of the PDict VIRTUAL class), a new container for storing a Trusted Band PDict object. - There are now 2 types of preprocessing: the "ACtree" type (the default) and the "Twobit" type. - Added the MTB_PDict class (another subclass of the PDict VIRTUAL class), a container for storing a Multiple Trusted Band PDict object. - The methods defined for PDict objects are now: length, width, names, [[, head, tb, tb.width, tail, show, duplicated and patternFrequency. - Changed the signature of the PDict() constructor: no more 'drop.head' and 'drop.tail' args, and new 'tb.width' and 'type' args. See ?PDict for the details (especially for the limitations of each type of preprocessing). STRING ALIGNMENT o Added support for character vectors of any length and XStringSet objects to the pattern argument of the pairwiseAlignment function. o Added "subjectOverlap" and "patternOverlap" pairwise sequence alignments. o Added support for Solexa quality scores in pairwise sequence alignment calculations. o Added support for fuzzy mappings in quality-based pairwise sequence alignments. o Added a stringDist function to calculate the Levenshtein edit distance between elements of a character vector or XStringSet. o Added many methods for pairwise alignment objects including as.matrix, compareStrings, consensusMatrix, consensusString, coverage, mismatchSummary, mismatchTable, nindel, nmatch, nmismatch, pattern, pid, rep, subject, summary, toString, Views. o Removed the XStringAlign class and added classes PairwiseAlignment, PairwiseAlignmentSummary, AlignedXStringSet, QualityAlignedXStringSet, QualityScaledXStringSet, QualityScaledBStringSet, QualityScaledDNAStringSet, QualityScaledRNAStringSet, QualityScaledAAStringSet, XStringQuality, PhredQuality, and SolexaQuality. MISCELLANEOUS ************************************************** * 2.8 SERIES NEWS * ************************************************** BASIC CONTAINERS o Added 2 containers for handling masked sequences: - The MaskCollection container for storing a collection of masks that can be used to mask regions in a sequence. - The MaskedXString family of containers for storing masked sequences. o Added new containers for storing a big set of sequences: - The XStringSet family: BStringSet, DNAStringSet, RNAStringSet and AAStringSet (all direct XStringSet subtypes with no additional slots). - The XStringList family: BStringList, DNAStringList, RNAStringList and AAStringList (all direct XStringList subtypes with no additional slots). The 2 families are almost the same from a user point of view, but the internal representations and method implementations are very different. The XStringList family was a first attempt to address the problem of storing a big set of sequences in an efficient manner but its performance turned out to be disappointing. So the XStringSet family was introduced as a response to the poor performance of the XStringList container. The XStringList family might be removed soon. o Added the trim() function for trimming the "out of limits" views of an XStringViews object. o Added "restrict", "narrow", "reduce" and "gaps" generic functions with methods for IRanges and XStringViews objects. These functions provide basic transformations of an IRanges object into another IRanges object of the same class. Also added the toNormalIRanges() function for normalizing an IRanges object. o Added the "start<-", "width<-" and "end<-" generics with methods for UnlockedIRanges and Views objects. Also added the "update" method for UnlockedIRanges objects to provide a convenient way of combining multiple modifications of an UnlockedIRanges object into one single call. o Added the intToRanges() and intToAdjacentRanges() utility functions for creating an IRanges instance. o Added the IRanges, UnlockedIRanges, Views, LockedIRanges and NormalIRanges classes for representing a set of integer ranges + the "isNormal" and "whichFirstNotNormal" generic functions with methods for IRanges objects (see ?IRanges for the details). Changed the definition of the XStringViews class so now it derives from the Views class. o Versatile constructor RNAString() (resp. DNAString()) now converts from DNA to RNA (resp. RNA to DNA) by replacing T by U (resp. U by T) instead of trying to mimic transcription. This conversion is still performed without copying the sequence data and thus remains very fast. Also the semantic of comparing RNA with DNA has been changed to remain consistent with the new semantic of RNAString() and DNAString() e.g. RNAString("UUGAAAA-CUC-N") is considered equal to DNAString("TTGAAAA-CTC-N"). o Added support for empty XString objects. o Added the XString() versatile constructor (it's a generic function with methods for character and XString objects). The BString(), DNAString(), RNAString() and AAString() constructors are now based on it. o Renamed subBString() -> subXString() and deprecated subBString(). o Renamed the BStringViews class -> XStringViews. o Reorganized the hierarchy of the BString class and subclasses by adding the XString virtual class: now the BString, DNAString, RNAString and AAString classes are all direct XString subtypes with no additional slots. Most importantly, they are all at the same level in the new hierarchy i.e. DNAString, RNAString and AAString objects are NOT BString objects anymore. C-LEVEL FACILITIES o Started the Biostrings C interface (work-in-progress). See inst/include/Biostrings_interface.h for how to use it in your package. UTILITIES o Added "reverse" methods for IRanges, NormalIRanges, MaskCollection and MaskedXString objects, and "complement" and "reverseComplement" methods for MaskedDNAString and MaskedRNAString objects. o Added the coverage() generic function with methods for IRanges, MaskCollection, XStringViews, MaskedXString and MIndex objects. o Added the injectHardMask() generic function for "hard masking" a sequence. o Added the maskMotif() generic function for masking a sequence by content. o Added utility functions for importing a mask: - read.agpMask(): read mask from an NCBI "agp" file; - read.gapMask(): read mask from an UCSC "gap" file; - read.liftMask(): read mask from an UCSC "lift" file; - read.rmMask(): read mask from a RepeatMasker .out file; - read.trfMask(): read mask from a Tandem Repeats Finder .bed file. o Added the subseq() generic function with methods for XString and MaskedXString objects. o Added functions read.BStringSet(), read.DNAStringSet(), read.RNAStringSet(), read.AAStringSet() and write.XStringSet(). read.BStringSet() and family is now preferred over read.XStringViews() for loading a FASTA file into R. Renamed helper function BStringViewsToFASTArecords() -> XStringSetToFASTArecords(). o Added the replaceLetterAtLoc() generic function with a method for DNAString objects (methods for other types of objects might come later) for making a copy of a sequence where letters are replaced by new letters at some specified locations. o Added the chartr() generic function with methods for XString, XStringSet and XStringViews objects. o Made the "show" methods for XString, XStringViews and XStringAlign objects "getOption('width') aware" so that the user can control the width of the output they produce. o Added the dinucleotideFrequency(), trinucleotideFrequency(), oligonucleotideFrequency(), strrev() and mkAllStrings() functions. o Four changes in alphabetFrequency(): (1) when used with 'baseOnly=TRUE', the frequency of the gap letter ("-") is not returned anymore (now it's treated as any 'other' letter i.e. any non-base letter); (2) added the 'freq' argument; (3) added the 'collapse' argument; (4) made it 1000x faster on XStringSet and XStringViews objects. o Added "as.character" and "consmat" methods for XStringAlign objects. o Added the patternFrequency() generic function with a method for CWdna_PDict objects (will come later for TBdna_PDict objects). o Added a "duplicated" method for CWdna_PDict objects (will come later for TBdna_PDict objects). o Added "reverse" method for XStringSet objects, and "complement" and "reverseComplement" methods for DNAStringSet and RNAStringSet objects. They all preserve the names. o reverse(), complement() and reverseComplement() now preserve the names when applied to an XStringViews object. o By Robert: Added the dna2rna(), rna2dna(), transcribe() and cDNA() functions + a "reverseComplement" method for RNAString objects. o Added the mergeIUPACLetters() utility function. STRING MATCHING o matchPattern.Rnw vignette replaced by much improved GenomeSearching.Rnw vignette (still a work-in-progress). o Added "matchPDict" methods for XStringViews and MaskedXString objects (only for a DNA input sequence). o Added support in matchPDict() for IUPAC ambiguities in the subject i.e. it will treat them as wildcards when called with 'fixed=FALSE' on a Trusted Band dict or with 'fixed=c(pattern=TRUE, subject=FALSE)' on any dict. o Added support in matchPDict() for inexact matching of a dictionary with "trusted prefixes". See ?`matchPDict-inexact` for the details. o Implemented the "shortcut feature" to C function CWdna_exact_search(). With this patch, using matchPDict() to find all the matches of a 3.3M 32-mers dictionary in the full Human genome (+ and - strands of all chromosomes) is about 2.5x faster than before (will take between 20 minutes and 2 hours depending on your machine and the number of matches found). This puts matchPDict() at the same level as the Vmatch software (http://www.vmatch.de/) for a dictionary of this size. Memory footprint for matchPDict() is about 2GB for the Aho-Corasick tree built from the 3.3M 32-mers dictionary. Building this tree is still very fast (2 or 3 minutes) (Vmatch needs 60G of disk space to build all its suffix arrays, don't know how long it takes for this, don't know what's the memory footprint either when they are loaded into memory but it looks like it is several gigabytes). matchPDict() only works with a dictionary of DNA patterns where all the patterns have the same number of nucleotides and it does only exact matching for now (Vmatch doesn't have this kind of limitations). o matchPDict() now returns an MIndex object (new class) instead of a list of integer vectors. The user can then extract the starts or the ends of the matches with startIndex() or endIndex(), extract the number of matches per pattern with countIndex(), extract the matches for a given pattern with [[, put all the matches in a single IRanges object with unlist() or convert this MIndex object into a set of views on the original subject with extractAllMatches(). Other functions can be added later in order to provide a wider choice of extraction/conversion tools if necessary. WARNING: This is still a work-in-progress. Function names and semantics are not yet stabilized! o Added the matchPDict() and countPDict() functions for efficiently finding (or just counting) all occurrences in a text (the subject) of any pattern from a set of patterns (the dictionary). The types of pattern dictionaries currently supported are constant width DNA dictionaries (CWdna_PDict objects) and "Trusted Prefix" DNA dictionaries (a particular case of "Trusted Band" DNA dictionaries, represented by TBdna_PDict objects). See ?matchPDict for the details (especially the current limitations). o Added basic support for palindrome finding: it can be achieved with the new findPalindromes() and findComplementedPalindromes() functions. Also added related utility functions palindromeArmLength(), palindromeLeftArm(), palindromeRightArm(), complementedPalindromeArmLength(), complementedPalindromeLeftArm() and complementedPalindromeRightArm(). o Added basic support for Position Weight Matrix matching thru the new matchPWM() and countPWM() functions. Also added related utility functions maxWeights(), maxScore() and PWMscore(). o Added "matchLRPatterns" and "matchProbePair" methods for XStringViews objects. o Added the nmismatchStartingAt(), nmismatchEndingAt() and isMatching() functions. o Change in terminology to align with established practices: "fuzzy matching" is now called "inexact matching". This change mostly affects the documentation. The only place where it also affects the API is that now 'algo="naive-inexact"' must be used instead of 'algo="naive-fuzzy"' when calling the matchPattern() function or any other function that has the 'algo' argument. o Renamed the 'mismatch' arg -> 'max.mismatch' for the matchPattern(), matchLRPatterns() and matchPDict() functions. MISCELLANEOUS o Renamed some files in inst/extdata/ to use the same extension (.fa) for all FASTA files. o Renamed Exfiles/ folder as extdata/ and put back fastaEx in it (from Biostrings 1). o Changed license from LGPL to Artistic-2.0 ************************************************** * 2.6 SERIES NEWS * ************************************************** o Added the matchLRPatterns() function for finding in a sequence patterns that are defined by a left and a right part. See ?matchLRPatterns for the details.