-# Copyright (C) 2007-2011 Martin A. Hansen.
+# Copyright (C) 2007-2012 Martin A. Hansen.
# This program is free software; you can redistribute it and/or
# modify it under the terms of the GNU General Public License
# >>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>><<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<
-require 'maasha/digest'
-require 'maasha/patternmatcher'
-#require 'maasha/patscan'
+require 'maasha/bits'
+require 'maasha/seq/backtrack'
+require 'maasha/seq/digest'
+#require 'maasha/seq/patscan'
+require 'maasha/seq/patternmatcher'
+require 'maasha/seq/trim'
+require 'narray'
# Residue alphabets
DNA = %w[a t c g]
INDELS = %w[. - _ ~]
# Quality scores bases
-SCORE_PHRED = 33
-SCORE_ILLUMINA = 64
+SCORE_BASE = 64
+SCORE_MIN = 0
+SCORE_MAX = 40
# Error class for all exceptions to do with Seq.
class SeqError < StandardError; end
class Seq
#include Patscan
include PatternMatcher
+ include BackTrack
+ include Digest
+ include Trim
attr_accessor :seq_name, :seq, :type, :qual
- # Method that generates all possible oligos of a specifed length and type.
+ # Class method to instantiate a new Sequence object given
+ # a Biopiece record.
+ def self.new_bp(record)
+ seq_name = record[:SEQ_NAME]
+ seq = record[:SEQ]
+ type = record[:SEQ_TYPE]
+ qual = record[:SCORES]
+
+ self.new(seq_name, seq, type, qual)
+ end
+
+ # Class method that generates all possible oligos of a specifed length and type.
def self.generate_oligos(length, type)
raise SeqError, "Cannot generate negative oligo length: #{length}" if length <= 0
self.seq.scan(regex).size
end
+ # Method to remove indels from seq and qual if qual.
+ def indels_remove
+ if self.qual.nil?
+ self.seq.delete!(Regexp.escape(INDELS.join('')))
+ else
+ na_seq = NArray.to_na(self.seq, "byte")
+ na_qual = NArray.to_na(self.qual, "byte")
+ mask = NArray.byte(self.length)
+
+ INDELS.each do |c|
+ mask += na_seq.eq(c.ord)
+ end
+
+ mask = mask.eq(0)
+
+ self.seq = na_seq[mask].to_s
+ self.qual = na_qual[mask].to_s
+ end
+
+ self
+ end
+
# Method that returns true is a given sequence type is DNA.
def is_dna?
self.type == 'dna'
# Method that given a Seq entry returns a FASTA entry (a string).
def to_fasta(wrap = nil)
- raise SeqError, "Missing seq_name" if self.seq_name.nil?
- raise SeqError, "Missing seq" if self.seq.nil?
+ raise SeqError, "Missing seq_name" if self.seq_name.nil? or self.seq_name == ''
+ raise SeqError, "Missing seq" if self.seq.nil? or self.seq.empty?
- seq_name = self.seq_name
- seq = self.seq
+ seq_name = self.seq_name.to_s
+ seq = self.seq.to_s
unless wrap.nil?
seq.gsub!(/(.{#{wrap}})/) do |match|
- match << "\n"
+ match << $/
end
seq.chomp!
end
- ">#{seq_name}\n#{seq}\n"
+ ">" + seq_name + $/ + seq + $/
+ end
+
+ # Method that given a Seq entry returns a FASTQ entry (a string).
+ def to_fastq
+ raise SeqError, "Missing seq_name" if self.seq_name.nil?
+ raise SeqError, "Missing seq" if self.seq.nil?
+ raise SeqError, "Missing qual" if self.qual.nil?
+
+ seq_name = self.seq_name.to_s
+ seq = self.seq.to_s
+ qual = self.qual.to_s
+
+ "@" + seq_name + $/ + seq + $/ + "+" + $/ + qual + $/
end
# Method that generates a unique key for a
def reverse_complement
self.reverse
self.complement
+ self
end
alias :revcomp :reverse_complement
# Method to reverse the sequence.
def reverse
self.seq.reverse!
+ self.qual.reverse! if self.qual
+ self
end
# Method that complements sequence including ambiguity codes.
raise SeqError, "Cannot complement 0 length sequence" if self.length == 0
if self.is_dna?
- self.seq.tr!( 'AGCUTRYWSMKHDVBNagcutrywsmkhdvbn', 'TCGAAYRWSKMDHBVNtcgaayrwskmdhbvn' )
+ self.seq.tr!('AGCUTRYWSMKHDVBNagcutrywsmkhdvbn', 'TCGAAYRWSKMDHBVNtcgaayrwskmdhbvn')
elsif self.is_rna?
- self.seq.tr!( 'AGCUTRYWSMKHDVBNagcutrywsmkhdvbn', 'UCGAAYRWSKMDHBVNucgaayrwskmdhbvn' )
+ self.seq.tr!('AGCUTRYWSMKHDVBNagcutrywsmkhdvbn', 'UCGAAYRWSKMDHBVNucgaayrwskmdhbvn')
else
raise SeqError, "Cannot complement sequence type: #{self.type}"
end
end
+ # Method to determine the Hamming Distance between
+ # two Sequence objects (case insensitive).
+ def hamming_distance(seq)
+ self.seq.upcase.hamming_distance(seq.seq.upcase)
+ end
+
# Method that generates a random sequence of a given length and type.
def generate(length, type)
raise SeqError, "Cannot generate sequence length < 1: #{length}" if length <= 0
seq_new
end
+ # Method to shuffle a sequence readomly inline.
+ def shuffle!
+ self.seq = self.seq.split('').shuffle!.join
+ self
+ end
+
# Method that returns a subsequence of from a given start position
# and of a given length.
def subseq(start, length = self.length - start)
raise SeqError, "subsequence start: #{start} < 0" if start < 0
- raise SeqError, "subsequence length: #{length} < 1" if length <= 0
+ raise SeqError, "subsequence length: #{length} < 0" if length < 0
raise SeqError, "subsequence start + length > Seq.length: #{start} + #{length} > #{self.length}" if start + length > self.length
- stop = start + length - 1
+ if length == 0
+ seq = ""
+ qual = "" unless self.qual.nil?
+ else
+ stop = start + length - 1
- seq = self.seq[start .. stop]
- qual = self.qual[start .. stop] unless self.qual.nil?
+ seq = self.seq[start .. stop]
+ qual = self.qual[start .. stop] unless self.qual.nil?
+ end
- Seq.new(self.seq_name, seq, self.type, qual)
+ Seq.new(self.seq_name.dup, seq, self.type, qual)
end
# Method that replaces a sequence with a subsequence from a given start position
# and of a given length.
def subseq!(start, length = self.length - start)
raise SeqError, "subsequence start: #{start} < 0" if start < 0
- raise SeqError, "subsequence length: #{length} < 1" if length <= 0
+ raise SeqError, "subsequence length: #{length} < 0" if length < 0
raise SeqError, "subsequence start + length > Seq.length: #{start} + #{length} > #{self.length}" if start + length > self.length
- stop = start + length - 1
+ if length == 0
+ self.seq = ""
+ self.qual = "" unless self.qual.nil?
+ else
+ stop = start + length - 1
+
+ self.seq = self.seq[start .. stop]
+ self.qual = self.qual[start .. stop] unless self.qual.nil?
+ end
- self.seq = self.seq[start .. stop]
- self.qual = self.qual[start .. stop] unless self.qual.nil?
+ self
end
# Method that returns a subsequence of a given length
((self.seq.scan(/[a-z]/).size.to_f / (self.len - self.indels).to_f) * 100).round(2)
end
- # Method to convert the quality scores from a specified base
- # to another base.
- def convert_phred2illumina!
- self.qual.gsub!(/./) do |score|
- score_phred = score.ord - SCORE_PHRED
- raise SeqError, "Bad Phred score: #{score} (#{score_phred})" unless (0 .. 41).include? score_phred
- score_illumina = score_phred + SCORE_ILLUMINA
- score = score_illumina.chr
+ # Hard masks sequence residues where the corresponding quality score
+ # is below a given cutoff.
+ def mask_seq_hard_old(cutoff)
+ seq = self.seq.upcase
+ scores = self.qual
+ i = 0
+
+ scores.each_char do |score|
+ seq[i] = 'N' if score.ord - SCORE_BASE < cutoff
+ i += 1
end
+
+ self.seq = seq
end
- # Method to convert the quality scores from Solexa odd/ratio to
- # Illumina format.
- def convert_solexa2illumina!
- self.qual.gsub!(/./) do |score|
- score = solexa_char2illumina_char(score)
+ # Hard masks sequence residues where the corresponding quality score
+ # is below a given cutoff.
+ def mask_seq_hard!(cutoff)
+ raise SeqError, "seq is nil" if self.seq.nil?
+ raise SeqError, "qual is nil" if self.qual.nil?
+ raise SeqError, "cufoff value: #{cutoff} out of range #{SCORE_MIN} .. #{SCORE_MAX}" unless (SCORE_MIN .. SCORE_MAX).include? cutoff
+
+ na_seq = NArray.to_na(self.seq, "byte")
+ na_qual = NArray.to_na(self.qual, "byte")
+ mask = (na_qual - SCORE_BASE) < cutoff
+ mask *= na_seq.ne("-".ord)
+
+ na_seq[mask] = 'N'.ord
+
+ self.seq = na_seq.to_s
+
+ self
+ end
+
+ # Soft masks sequence residues where the corresponding quality score
+ # is below a given cutoff.
+ def mask_seq_soft!(cutoff)
+ raise SeqError, "seq is nil" if self.seq.nil?
+ raise SeqError, "qual is nil" if self.qual.nil?
+ raise SeqError, "cufoff value: #{cutoff} out of range #{SCORE_MIN} .. #{SCORE_MAX}" unless (SCORE_MIN .. SCORE_MAX).include? cutoff
+
+ na_seq = NArray.to_na(self.seq, "byte")
+ na_qual = NArray.to_na(self.qual, "byte")
+ mask = (na_qual - SCORE_BASE) < cutoff
+ mask *= na_seq.ne("-".ord)
+
+ na_seq[mask] ^= ' '.ord
+
+ self.seq = na_seq.to_s
+
+ self
+ end
+
+ # Method to convert quality scores inbetween formats.
+ # Sanger base 33, range 0-40
+ # Solexa base 64, range -5-40
+ # Illumina13 base 64, range 0-40
+ # Illumina15 base 64, range 3-40
+ # Illumina18 base 33, range 0-41
+ def convert_scores!(from, to)
+ unless from == to
+ na_qual = NArray.to_na(self.qual, "byte")
+
+ case from.downcase
+ when "sanger" then na_qual -= 33
+ when "solexa" then na_qual -= 64
+ when "illumina13" then na_qual -= 64
+ when "illumina15" then na_qual -= 64
+ when "illumina18" then na_qual -= 33
+ else raise SeqError, "unknown quality score encoding: #{from}"
+ end
+
+ case to.downcase
+ when "sanger" then na_qual += 33
+ when "solexa" then na_qual += 64
+ when "illumina13" then na_qual += 64
+ when "illumina15" then na_qual += 64
+ when "illumina18" then na_qual += 33
+ else raise SeqError, "unknown quality score encoding: #{from}"
+ end
+
+ self.qual = na_qual.to_s
end
+
+ self
end
- private
+ # Method to calculate and return the mean quality score.
+ def scores_mean
+ raise SeqError, "Missing qual in entry" if self.qual.nil?
- # Method to convert a Solexa score (odd ratio) to
- # a phred (probability) integer score.
- def solexa2phred(score)
- (10.0 * Math.log(10.0 ** (score / 10.0) + 1.0, 10)).to_i
+ na_qual = NArray.to_na(self.qual, "byte")
+ na_qual -= SCORE_BASE
+ na_qual.mean
end
- # Method to convert a Solexa score encoded using base
- # 64 ASCII to a Phred score encoded using base 64 ASCII.
- def solexa_char2illumina_char(char)
- score_solexa = char.ord - 64
- score_phred = solexa2phred(score_solexa)
- (score_phred + 64).chr
+ # Method to find open reading frames (ORFs).
+ def each_orf(size_min, size_max, start_codons, stop_codons, pick_longest = false)
+ orfs = []
+ pos_beg = 0
+
+ regex_start = Regexp.new(start_codons.join('|'), true)
+ regex_stop = Regexp.new(stop_codons.join('|'), true)
+
+ while pos_beg and pos_beg < self.length - size_min
+ if pos_beg = self.seq.index(regex_start, pos_beg)
+ if pos_end = self.seq.index(regex_stop, pos_beg)
+ length = (pos_end - pos_beg) + 3
+
+ if (length % 3) == 0
+ if size_min <= length and length <= size_max
+ subseq = self.subseq(pos_beg, length)
+
+ orfs << [subseq, pos_beg, pos_end + 3]
+ end
+ end
+ end
+
+ pos_beg += 1
+ end
+ end
+
+ if pick_longest
+ orf_hash = {}
+
+ orfs.each { |orf| orf_hash[orf.last] = orf unless orf_hash[orf.last] }
+
+ orfs = orf_hash.values
+ end
+
+ if block_given?
+ orfs.each { |orf| yield orf }
+ else
+ return orfs
+ end
end
end
projects / biopieces.git / blobdiff