--- /dev/null	Thu Jan 01 00:00:00 1970 +0000
+++ b/mismatch_frequencies.py	Tue Jan 27 05:55:52 2015 -0500
@@ -0,0 +1,229 @@
+import pysam, re, string
+import matplotlib.pyplot as plt
+import pandas as pd
+from collections import defaultdict
+from collections import OrderedDict
+import argparse
+
+class MismatchFrequencies:
+    '''Iterate over a SAM/BAM alignment file, collecting reads with mismatches. One
+    class instance per alignment file. The result_dict attribute will contain a
+    nested dictionary with name, readlength and mismatch count.'''
+    def __init__(self, result_dict={}, alignment_file=None, name="name", minimal_readlength=21, maximal_readlength=21,
+                 number_of_allowed_mismatches=1, ignore_5p_nucleotides=0, ignore_3p_nucleotides=0):
+    
+        self.result_dict = result_dict
+        self.name = name
+        self.minimal_readlength = minimal_readlength
+        self.maximal_readlength = maximal_readlength
+        self.number_of_allowed_mismatches = number_of_allowed_mismatches
+        self.ignore_5p_nucleotides = ignore_5p_nucleotides
+        self.ignore_3p_nucleotides = ignore_3p_nucleotides
+        
+        if alignment_file:
+            self.pysam_alignment = pysam.Samfile(alignment_file)
+            result_dict[name]=self.get_mismatches(self.pysam_alignment, minimal_readlength, maximal_readlength)
+    
+    def get_mismatches(self, pysam_alignment, minimal_readlength, maximal_readlength):
+        mismatch_dict = defaultdict(int)
+        len_dict={}
+        for i in range(minimal_readlength, maximal_readlength+1):
+            len_dict[i]=mismatch_dict.copy()
+        for alignedread in pysam_alignment:
+            if self.read_is_valid(alignedread, minimal_readlength, maximal_readlength):
+                len_dict[int(alignedread.rlen)]['total_mapped'] += 1
+                MD=alignedread.opt('MD')
+                if self.read_has_mismatch(alignedread, self.number_of_allowed_mismatches):
+                    (ref_base, mismatch_base)=self.read_to_reference_mismatch(MD, alignedread.seq, alignedread.is_reverse)
+                    if ref_base == None:
+                            continue
+                    else:
+                        for i, base in enumerate(ref_base):
+                            len_dict[int(alignedread.rlen)][ref_base[i]+' to '+mismatch_base[i]] += 1
+        return len_dict
+    
+    def read_is_valid(self, read, min_readlength, max_readlength):
+        '''Filter out reads that are unmatched, too short or
+        too long or that contian insertions'''
+        if read.is_unmapped:
+            return False
+        if read.rlen < min_readlength:
+            return False
+        if read.rlen > max_readlength:
+            return False
+        else:
+            return True
+    
+    def read_has_mismatch(self, read, number_of_allowed_mismatches=1):
+        '''keep only reads with one mismatch. Could be simplified'''
+        NM=read.opt('NM')
+        if NM <1: #filter out reads with no mismatch
+            return False
+        if NM >number_of_allowed_mismatches: #filter out reads with more than 1 mismtach
+            return False
+        else:
+            return True
+        
+    def mismatch_in_allowed_region(self, readseq, mismatch_position):
+        '''
+        >>> M = MismatchFrequencies()
+        >>> readseq = 'AAAAAA'
+        >>> mismatch_position = 2
+        >>> M.mismatch_in_allowed_region(readseq, mismatch_position)
+        True
+        >>> M = MismatchFrequencies(ignore_3p_nucleotides=2, ignore_5p_nucleotides=2)
+        >>> readseq = 'AAAAAA'
+        >>> mismatch_position = 1
+        >>> M.mismatch_in_allowed_region(readseq, mismatch_position)
+        False
+        >>> readseq = 'AAAAAA'
+        >>> mismatch_position = 4
+        >>> M.mismatch_in_allowed_region(readseq, mismatch_position)
+        False
+        '''
+        mismatch_position+=1 # To compensate for starting the count at 0
+        five_p = self.ignore_5p_nucleotides
+        three_p = self.ignore_3p_nucleotides
+        if any([five_p > 0, three_p > 0]):
+            if any([mismatch_position <= five_p, 
+                    mismatch_position >= (len(readseq)+1-three_p)]): #Again compensate for starting the count at 0
+                return False
+            else:
+                return True
+        else:
+            return True
+            
+    def read_to_reference_mismatch(self, MD, readseq, is_reverse):
+        '''
+        This is where the magic happens. The MD tag contains SNP and indel information,
+        without looking to the genome sequence. This is a typical MD tag: 3C0G2A6.
+        3 bases of the read align to the reference, followed by a mismatch, where the
+        reference base is C, followed by 10 bases aligned to the reference. 
+        suppose a reference 'CTTCGATAATCCTT'
+                             |||  || ||||||
+                 and a read 'CTTATATTATCCTT'. 
+        This situation is represented by the above MD tag. 
+        Given MD tag and read sequence this function returns the reference base C, G and A, 
+        and the mismatched base A, T, T.
+        >>> M = MismatchFrequencies()
+        >>> MD='3C0G2A7'
+        >>> seq='CTTATATTATCCTT'
+        >>> result=M.read_to_reference_mismatch(MD, seq, is_reverse=False)
+        >>> result[0]=="CGA"
+        True
+        >>> result[1]=="ATT"
+        True
+        >>> 
+        '''
+        search=re.finditer('[ATGC]',MD)
+        if '^' in MD:
+            print 'WARNING insertion detected, mismatch calling skipped for this read!!!'
+            return (None, None)
+        start_index=0 # refers to the leading integer of the MD string before an edited base
+        current_position=0 # position of the mismatched nucleotide in the MD tag string
+        mismatch_position=0 # position of edited base in current read 
+        reference_base=""
+        mismatched_base=""
+        for result in search:
+            current_position=result.start()
+            mismatch_position=mismatch_position+1+int(MD[start_index:current_position]) #converts the leading characters before an edited base into integers
+            start_index=result.end()
+            reference_base+=MD[result.end()-1]
+            mismatched_base+=readseq[mismatch_position-1]
+        if is_reverse:
+            reference_base=reverseComplement(reference_base)
+            mismatched_base=reverseComplement(mismatched_base)
+        if mismatched_base=='N':
+            return (None, None)
+        if self.mismatch_in_allowed_region(readseq, mismatch_position):
+            return (reference_base, mismatched_base)
+        else:
+            return (None, None)
+
+
+def reverseComplement(sequence):
+    '''do a reverse complement of DNA base.
+    >>> reverseComplement('ATGC')=='GCAT'
+    True
+    >>> 
+    '''
+    sequence=sequence.upper()
+    complement = string.maketrans('ATCGN', 'TAGCN')
+    return sequence.upper().translate(complement)[::-1]
+
+def barplot(df, library, axes):
+    df.plot(kind='bar', ax=axes, subplots=False,\
+            stacked=False, legend='test',\
+            title='Mismatches in TE small RNAs from {0}'.format(library))
+  
+def result_dict_to_df(result_dict):
+    mismatches = []
+    libraries = []
+    for mismatch, library in result_dict.iteritems():
+        mismatches.append(mismatch)
+        libraries.append(pd.DataFrame.from_dict(library, orient='index'))
+    df=pd.concat(libraries, keys=mismatches)
+    df.index.names = ['library', 'readsize']
+    return df
+
+def df_to_tab(df, output):
+    df.to_csv(output, sep='\t')
+
+def plot_result(result_dict, args):
+    names=args.name
+    nrows=len(names)/2+1
+    fig = plt.figure(figsize=(16,32))
+    for i,library in enumerate (names):
+        axes=fig.add_subplot(nrows,2,i+1)
+        library_dict=result_dict[library]
+        for length in library_dict.keys():
+            for mismatch in library_dict[length]:
+                if mismatch == 'total_mapped':
+                    continue
+                library_dict[length][mismatch]=library_dict[length][mismatch]/float(library_dict[length]['total_mapped'])*100
+            del library_dict[length]['total_mapped']
+        df=pd.DataFrame(library_dict)
+        barplot(df, library, axes),
+        axes.set_ylabel('Percent of mapped reads with mismatches')
+    fig.savefig(args.output_pdf, format='pdf')    
+
+def setup_MismatchFrequencies(args):
+    resultDict=OrderedDict()
+    kw_list=[{'result_dict' : resultDict, 
+             'alignment_file' :alignment_file, 
+             'name' : name, 
+             'minimal_readlength' : args.min, 
+             'maximal_readlength' : args.max,
+             'number_of_allowed_mismatches' : args.n_mm,
+             'ignore_5p_nucleotides' : args.five_p, 
+             'ignore_3p_nucleotides' : args.three_p} 
+             for alignment_file, name in zip(args.input, args.name)]
+    return (kw_list, resultDict)
+
+def run_MismatchFrequencies(args):
+    kw_list, resultDict=setup_MismatchFrequencies(args)
+    [MismatchFrequencies(**kw_dict) for kw_dict in kw_list]
+    return resultDict
+
+def main():
+    result_dict=run_MismatchFrequencies(args)
+    df=result_dict_to_df(result_dict)
+    plot_result(result_dict, args)
+    df_to_tab(df, args.output_tab)
+
+if __name__ == "__main__":
+    
+    parser = argparse.ArgumentParser(description='Produce mismatch statistics for BAM/SAM alignment files.')
+    parser.add_argument('--input', nargs='*', help='Input files in SAM/BAM format')
+    parser.add_argument('--name', nargs='*', help='Name for input file to display in output file. Should have same length as the number of inputs')
+    parser.add_argument('--output_pdf', help='Output filename for graph')
+    parser.add_argument('--output_tab', help='Output filename for table')
+    parser.add_argument('--min', '--minimal_readlength', type=int, help='minimum readlength')
+    parser.add_argument('--max', '--maximal_readlength', type=int, help='maximum readlength')
+    parser.add_argument('--n_mm', '--number_allowed_mismatches', type=int, default=1, help='discard reads with more than n mismatches')
+    parser.add_argument('--five_p', '--ignore_5p_nucleotides', type=int, default=0, help='when calculating nucleotide mismatch frequencies ignore the first N nucleotides of the read')
+    parser.add_argument('--three_p', '--ignore_3p_nucleotides', type=int, default=1, help='when calculating nucleotide mismatch frequencies ignore the last N nucleotides of the read')
+    #args = parser.parse_args(['--input', '3mismatches_ago2ip.bam', '2mismatch.bam', '--name', 'Siomi1', 'Siomi2' , '--five_p', '3','--three_p','3','--output_pdf', 'out.pdf', '--output_tab', 'out.tab', '--min', '21', '--max', '21'])
+    args = parser.parse_args()
+    main()
+