def write_intron_fasta(transcript_dict, fasta_dict, prefix='introns', sense=True):
seq_dict = {}
for transcript, values in transcript_dict.iteritems():
start = values[0]
end = values[1]
strand = values[2]
chrom = values[3]
CDS_start_list = values[4]
CDS_end_list = values[5]
for n in range(len(CDS_start_list)-1):
if strand == '+':
seq = fasta_dict[chrom][CDS_end_list[n]:CDS_start_list[n+1]-1]
elif strand == '-':
intron = len(CDS_start_list)-n-1
seq = fasta_dict[chrom][CDS_end_list[intron]:CDS_start_list[intron-1]-1]
seq = SP.reverse_complement(seq)
if sense is False:
seq = SP.reverse_complement(seq)
seq_dict[transcript+'_'+str(n)] = seq
with open('{}.fa'.format(prefix), 'w') as fout:
for transcript, seq in seq_dict.iteritems():
fout.write('>'+transcript+'\n')
fout.write(seq+'\n')
return seq_dict
def add_seq(branch_df, fa_dict):
five_seqs = []
branch_seqs = []
for ix, r in branch_df.iterrows():
five = fa_dict[r['chromosome']][r['5p splice site']-8:r['5p splice site']+8]
branch = fa_dict[r['chromosome']][r['branch site']-8:r['branch site']+8]
if r['strand'] == '-':
five = SP.reverse_complement(five)
branch = SP.reverse_complement(branch)
if 'GT' in five[4:11]:
ix = five.index('GT')
five = five[ix-2:ix+6]
else:
five = five[4:12]
if 'AG' in branch[4:11]:
ix = branch.index('AG')
branch = branch[ix-4:ix+4]
elif 'AA' in branch[4:11]:
ix = branch.index('AA')
branch = branch[ix-4:ix+4]
elif 'GA' in branch[4:11]:
ix = branch.index('GA')
branch = branch[ix-4:ix+4]
else:
branch = branch[4:13]
five_seqs.append(five)
branch_seqs.append(branch)
branch_df['5p seq'] = five_seqs
branch_df['Branch seq'] = branch_seqs
receptors = ['AG', 'AA', 'GA']
branch_df = branch_df[branch_df['Branch seq'].str[4:6].isin(receptors)]
return branch_df
def generate_all_ss_seqs(gff3, fasta_dict, organism):
transcript_dict = SP.build_transcript_dict(gff3, organism=organism)
ss, flag = SP.list_splice_sites(gff3, organism=organism)
ss_dict = SP.collapse_ss_dict(ss)
all_seq5 = []
all_seq3 = []
for transcript, introns in ss_dict.iteritems():
if organism == 'pombe':
isoform = transcript+'.1'
else:
isoform = transcript+'T0'
strand = transcript_dict[isoform][2]
chrom = transcript_dict[isoform][3]
for intron in introns:
if strand == '+':
seq5 = fasta_dict[chrom][(intron[0]-1):(intron[0]+7)]
elif strand == '-':
seq5 = fasta_dict[chrom][(intron[0]-6):(intron[0]+2)]
seq5 = SP.reverse_complement(seq5)
all_seq5.append(seq5)
if strand == '+':
seq3 = fasta_dict[chrom][(intron[1]-5):(intron[1]+3)]
elif strand == '-':
seq3 = fasta_dict[chrom][(intron[1]-2):(intron[1]+6)]
seq3 = SP.reverse_complement(seq3)
all_seq3.append(seq3)
return all_seq5, all_seq3
def add_seq(branch_df, fa_dict):
five_seqs = []
branch_seqs = []
for ix, r in branch_df.iterrows():
five = fa_dict[r['chromosome']][r['5p splice site'] -
8:r['5p splice site'] + 8]
branch = fa_dict[r['chromosome']][r['branch site'] -
8:r['branch site'] + 8]
if r['strand'] == '-':
five = SP.reverse_complement(five)
branch = SP.reverse_complement(branch)
if 'GT' in five[4:11]:
ix = five.index('GT')
five = five[ix - 2:ix + 6]
else:
five = five[4:12]
if 'AG' in branch[4:11]:
ix = branch.index('AG')
branch = branch[ix - 4:ix + 4]
elif 'AA' in branch[4:11]:
ix = branch.index('AA')
branch = branch[ix - 4:ix + 4]
elif 'GA' in branch[4:11]:
ix = branch.index('GA')
branch = branch[ix - 4:ix + 4]
else:
branch = branch[4:13]
five_seqs.append(five)
branch_seqs.append(branch)
branch_df['5p seq'] = five_seqs
branch_df['Branch seq'] = branch_seqs
receptors = ['AG', 'AA', 'GA']
branch_df = branch_df[branch_df['Branch seq'].str[4:6].isin(receptors)]
return branch_df
def generate_all_ss_seqs(gff3, fasta_dict, organism):
transcript_dict = SP.build_transcript_dict(gff3, organism=organism)
ss, flag = SP.list_splice_sites(gff3, organism=organism)
ss_dict = SP.collapse_ss_dict(ss)
all_seq5 = []
all_seq3 = []
for transcript, introns in ss_dict.iteritems():
if organism == 'pombe':
isoform = transcript + '.1'
else:
isoform = transcript + 'T0'
strand = transcript_dict[isoform][2]
chrom = transcript_dict[isoform][3]
for intron in introns:
if strand == '+':
seq5 = fasta_dict[chrom][(intron[0] - 1):(intron[0] + 7)]
elif strand == '-':
seq5 = fasta_dict[chrom][(intron[0] - 6):(intron[0] + 2)]
seq5 = SP.reverse_complement(seq5)
all_seq5.append(seq5)
if strand == '+':
seq3 = fasta_dict[chrom][(intron[1] - 5):(intron[1] + 3)]
elif strand == '-':
seq3 = fasta_dict[chrom][(intron[1] - 2):(intron[1] + 6)]
seq3 = SP.reverse_complement(seq3)
all_seq3.append(seq3)
return all_seq5, all_seq3
def write_intergenic_fasta(transcript_dict, fasta_dict, bps_us=0, bps_ds=0, all_intergenic=True, prefix='intergenic_transcripts'):
seq_dict = {}
if all_intergenic is False:
for transcript, values in transcript_dict.iteritems():
start = values[0]
end = values[1]
strand = values[2]
chrom = values[3]
if bps_us > 0:
if strand == '+':
seq_us_sense = fasta_dict[chrom][start-bps_us:start]
elif strand == '-':
seq_us_sense = fasta_dict[chrom][end:end+bps_us]
seq_us_sense = SP.reverse_complement(seq_us_sense)
seq_us_antisense = SP.reverse_complement(seq_us_sense)
seq_dict[transcript+'_us_sense'] = seq_us_sense
seq_dict[transcript+'_us_antisense'] = seq_us_antisense
if bps_ds > 0:
if strand == '+':
seq_ds_sense = fasta_dict[chrom][end:bps_ds+end]
elif strand == '-':
seq_ds_sense = fasta_dict[chrom][start-bps_ds:start]
seq_ds_sense = SP.reverse_complement(seq_ds_sense)
seq_ds_antisense = SP.reverse_complement(seq_ds_sense)
seq_dict[transcript+'_ds_sense'] = seq_ds_sense
seq_dict[transcript+'_ds_antisense'] = seq_ds_antisense
elif all_intergenic is True:
chroms = fasta_dict.keys()
for chrom in chroms:
chrom_transcripts = dict((k, transcript_dict[k]) for k in transcript_dict if transcript_dict[k][3] == chrom)
chr_txs_df = pd.DataFrame.from_dict(chrom_transcripts, orient='index')
chr_txs_df.sort_values([0], inplace=True)
sorted_transcripts = chr_txs_df.index.tolist()
n = 0
for n in range(len(sorted_transcripts)-1):
transcript = sorted_transcripts[n]
next_transcript = sorted_transcripts[n+1]
transcript_end = chr_txs_df[1][transcript]
next_start = chr_txs_df[0][next_transcript]
if next_start > transcript_end:
seq_plus = fasta_dict[chrom][transcript_end:next_start]
seq_dict[transcript+'_'+next_transcript+'_plus'] = seq_plus
seq_dict[transcript+'_'+next_transcript+'_minus'] = SP.reverse_complement(seq_plus)
else:
print 'Overlapping transcripts:'
print transcript
print next_transcript
with open('{}.fa'.format(prefix), 'w') as fout:
for transcript, seq in seq_dict.iteritems():
fout.write('>'+transcript+'\n')
fout.write(seq+'\n')
return seq_dict
def get_sequence(coord_dict, gff3_file, fasta_file):
if 'pombe' in gff3_file:
organism = 'pombe'
else:
organism = None
transcript_dict = SP.build_transcript_dict(gff3_file, organism=organism)
if type(fasta_file) is str:
fasta_dict = make_fasta_dict(fasta_file)
else:
fasta_dict = fasta_file
seq_dict = {}
counter5 = 0
counter3 = 0
other = 0
for transcript, coord_sets in coord_dict.iteritems():
seq_dict[transcript] = []
chrom = transcript_dict[transcript][3]
#if chrom in rom_lat: chrom = rom_lat[chrom]
strand = transcript_dict[transcript][2]
for coord in coord_sets[0]:
seq_type = 'other'
if strand == "+":
sequence = fasta_dict[chrom][(coord - 9):(coord + 11)]
elif strand == "-":
sequence = fasta_dict[chrom][(coord - 10):(coord + 10)]
sequence = SP.reverse_complement(sequence)
if sequence[10:12] == 'GT' or sequence[10:12] == 'GC':
seq_type = "5'"
counter5 += 1
seq_dict[transcript].append((sequence, seq_type))
for coord in coord_sets[1]:
seq_type = 'other'
if strand == "+":
sequence = fasta_dict[chrom][(coord - 9):(coord + 11)]
elif strand == "-":
sequence = fasta_dict[chrom][(coord - 10):(coord + 10)]
sequence = SP.reverse_complement(sequence)
if sequence[8:10] == 'AG':
seq_type = "3'"
counter3 += 1
seq_dict[transcript].append((sequence, seq_type))
#print str(counter5)+" 5' splice sites"
#print str(counter3)+" 3' splice sites"
return seq_dict
def get_sequence(coord_dict, gff3_file, fasta_file):
if 'pombe' in gff3_file:
organism = 'pombe'
else: organism = None
transcript_dict = SP.build_transcript_dict(gff3_file, organism=organism)
if type(fasta_file) is str:
fasta_dict = make_fasta_dict(fasta_file)
else:
fasta_dict = fasta_file
seq_dict = {}
counter5 = 0
counter3 = 0
other = 0
for transcript, coord_sets in coord_dict.iteritems():
seq_dict[transcript] = []
chrom = transcript_dict[transcript][3]
#if chrom in rom_lat: chrom = rom_lat[chrom]
strand = transcript_dict[transcript][2]
for coord in coord_sets[0]:
seq_type = 'other'
if strand == "+":
sequence = fasta_dict[chrom][(coord-9):(coord+11)]
elif strand == "-":
sequence = fasta_dict[chrom][(coord-10):(coord+10)]
sequence = SP.reverse_complement(sequence)
if sequence[10:12] == 'GT' or sequence[10:12] == 'GC':
seq_type = "5'"
counter5 += 1
seq_dict[transcript].append((sequence, seq_type))
for coord in coord_sets[1]:
seq_type = 'other'
if strand == "+":
sequence = fasta_dict[chrom][(coord-9):(coord+11)]
elif strand == "-":
sequence = fasta_dict[chrom][(coord-10):(coord+10)]
sequence = SP.reverse_complement(sequence)
if sequence[8:10] == 'AG':
seq_type = "3'"
counter3 += 1
seq_dict[transcript].append((sequence, seq_type))
#print str(counter5)+" 5' splice sites"
#print str(counter3)+" 3' splice sites"
return seq_dict
def seq_simple(chrom, start, end, strand, fasta_dict):
if type(fasta_dict) == str:
with open(fasta_dict, 'r') as f:
fasta_dict = json.load(f)
seq = fasta_dict[chrom][start:end+1]
if strand == '-':
seq = SP.reverse_complement(seq)
return seq
def seq_simple(chrom, start, end, strand, fasta_dict):
if type(fasta_dict) == str:
with open(fasta_dict, 'r') as f:
fasta_dict = json.load(f)
seq = fasta_dict[chrom][start:end + 1]
if strand == '-':
seq = SP.reverse_complement(seq)
return seq
def score_PyTract(df, fa_dict, alt_column_name=None, from_branches=False):
py_score1 = []
py_score2 = []
alt_py1 = []
alt_py2 = []
for ix, r in df.iterrows():
strand = r['strand']
chrom = r['chromosome']
coord = r['annotated intron coords'][1]
alt_coord = r['junction coords'][1]
if strand == '+':
if coord is not None:
seq1 = fa_dict[chrom][coord - 15:coord]
seq2 = fa_dict[chrom][coord - 30:coord - 15]
alt1 = fa_dict[chrom][alt_coord - 15:alt_coord]
alt2 = fa_dict[chrom][alt_coord - 30:alt_coord - 15]
if strand == '-':
if coord is not None:
seq1 = fa_dict[chrom][coord:coord + 15]
seq2 = fa_dict[chrom][coord + 15:coord + 30]
seq1 = SP.reverse_complement(seq1)
seq2 = SP.reverse_complement(seq2)
alt1 = fa_dict[chrom][alt_coord:alt_coord + 15]
alt2 = fa_dict[chrom][alt_coord + 15:alt_coord + 30]
alt1 = SP.reverse_complement(alt1)
alt2 = SP.reverse_complement(alt2)
alt_py1.append(percent_py(alt1))
alt_py2.append(percent_py(alt2))
if coord is not None:
py_score1.append(percent_py(seq1))
py_score2.append(percent_py(seq2))
else:
py_score1.append(np.NaN)
py_score2.append(np.NaN)
df['Py score annotated -15:0'] = py_score1
df['Py score annotated -30:-15'] = py_score2
df['Py score alternative -15:0'] = alt_py1
df['Py score alternative -30:-15'] = alt_py2
return df
def score_PyTract(df, fa_dict, alt_column_name=None, from_branches=False):
py_score1 = []
py_score2 = []
alt_py1 = []
alt_py2 = []
for ix, r in df.iterrows():
strand = r['strand']
chrom = r['chromosome']
coord = r['annotated intron coords'][1]
alt_coord = r['junction coords'][1]
if strand == '+':
if coord is not None:
seq1 = fa_dict[chrom][coord-15:coord]
seq2 = fa_dict[chrom][coord-30:coord-15]
alt1 = fa_dict[chrom][alt_coord-15:alt_coord]
alt2 = fa_dict[chrom][alt_coord-30:alt_coord-15]
if strand == '-':
if coord is not None:
seq1 = fa_dict[chrom][coord:coord+15]
seq2 = fa_dict[chrom][coord+15:coord+30]
seq1 = SP.reverse_complement(seq1)
seq2 = SP.reverse_complement(seq2)
alt1 = fa_dict[chrom][alt_coord:alt_coord+15]
alt2 = fa_dict[chrom][alt_coord+15:alt_coord+30]
alt1 = SP.reverse_complement(alt1)
alt2 = SP.reverse_complement(alt2)
alt_py1.append(percent_py(alt1))
alt_py2.append(percent_py(alt2))
if coord is not None:
py_score1.append(percent_py(seq1))
py_score2.append(percent_py(seq2))
else:
py_score1.append(np.NaN)
py_score2.append(np.NaN)
df['Py score annotated -15:0'] = py_score1
df['Py score annotated -30:-15'] = py_score2
df['Py score alternative -15:0'] = alt_py1
df['Py score alternative -30:-15'] = alt_py2
return df
def write_transcript_fasta(transcript_dict, fasta_dict, prefix='transcripts', sense=True, spliced=False):
seq_dict = {}
for transcript, values in transcript_dict.iteritems():
start = values[0]
end = values[1]
strand = values[2]
chrom = values[3]
CDS_start_list = values[4]
CDS_end_list = values[5]
if spliced is False:
seq = fasta_dict[chrom][start-1:end]
if strand == '-':
seq = SP.reverse_complement(seq)
elif spliced is True:
seq = ''
for n in range(len(CDS_start_list)):
if strand == '+':
seq = seq+fasta_dict[chrom][CDS_start_list[n]-1:CDS_end_list[n]]
elif strand == '-':
new_seq = fasta_dict[chrom][CDS_start_list[n]-1:CDS_end_list[n]]
new_seq = SP.reverse_complement(new_seq)
seq = seq+new_seq
if sense is False:
seq = SP.reverse_complement(seq)
seq_dict[transcript] = seq
with open('{}.fa'.format(prefix), 'w') as fout:
for transcript, seq in seq_dict.iteritems():
fout.write('>'+transcript+'\n')
fout.write(seq+'\n')
return seq_dict
def collect_intron_seq(gff3_file, fasta_file, ss_dict=None, junction_bed=None, gene_list=None, peak_df=None, organism=None):
transcript_dict = SP.build_transcript_dict(gff3_file, organism=organism)
if type(fasta_file) == dict:
fasta_dict = fasta_file
elif fasta_file.endswith('json'):
with open(fasta_file, 'r') as f:
fasta_dict = json.load(f)
else:
fasta_dict = make_fasta_dict(fasta_file)
if ss_dict is not None:
ss_dict=ss_dict
elif junction_bed is not None:
ss_dict = SP.build_junction_dict(junction_bed, gff3_file, transcript_dict, organism=organism)
elif peak_df is not None:
ss_dict = {}
peak_df = peak_df[~peak_df['type'].str.contains('prime')]
for ix, r in peak_df.iterrows():
if r['transcript'] not in ss_dict:
ss_dict[r['transcript']] = []
if r['strand'] == '+':
ss_dict[r['transcript']].append((r['position'],r['position']+50))
elif r['strand'] == '-':
ss_dict[r['transcript']].append((r['position'],r['position']-50))
else:
ss_dict, intron_flag = SP.list_splice_sites(gff3_file, gene_list=gene_list, organism=organism)
ss_dict = SP.collapse_ss_dict(ss_dict)
seq_dict = {}
for transcript, introns in ss_dict.iteritems():
if junction_bed is None:
if organism == 'pombe':
transcript = transcript+'.1'
else:
transcript = transcript+'T0'
introns = list(introns)
strand = transcript_dict[transcript][2]
chrom = transcript_dict[transcript][3]
n = 0
for n in range(len(introns)):
if strand == '+':
seq_dict[transcript+'-'+chrom+':'+str(introns[n][0]+1)] = fasta_dict[chrom][introns[n][0]+2:introns[n][0]+17]
elif strand == '-':
seq = fasta_dict[chrom][introns[n][0]-16:introns[n][0]-1]
seq_dict[transcript+'-'+chrom+':'+str(introns[n][0])] = SP.reverse_complement(seq)
return seq_dict
def backfill_splice_sites(df, gff3, fa_dict, PSSM, organism=None):
tx_dict = SP.build_transcript_dict(gff3, organism=organism)
ss_dict, flag = SP.list_splice_sites(gff3, organism=organism)
ss_dict = SP.collapse_ss_dict(ss_dict)
column_dict = {'position':[],'transcript':[],'alt splicing':[],'type':[],'strand':[], 'introns in transcript':[],
'intron size':[],'chromosome':[], '5p score':[], '3p score':[], 'intron position':[], 'exon size (us)':[],
'exon size (ds)':[],'transcript size':[], 'peak':[], 'seq5':[],'seq3':[]}
new_index = []
for tx in set(df['transcript']):
strand = df[df['transcript'] == tx].iloc[0]['strand']
splice_sites = ss_dict[tx]
if strand == '+':
splice_sites = sorted(list(splice_sites), key=lambda x:x[0])
elif strand == '-':
splice_sites = sorted(list(splice_sites), key=lambda x:x[0], reverse=True)
df_pos = None
for n, (five_site, three_site) in enumerate(splice_sites):
# Check if already in dataframe
in_df = False
for peak in df[df['transcript'] == tx]['position']:
if five_site in range(int(peak)-5,int(peak)+5):
in_df = True
df_pos = peak
break
column_dict['transcript'].append(tx)
if organism == 'pombe':
iso = tx+'.1'
else: iso = tx+'T0'
column_dict['intron size'].append(abs(three_site-five_site))
column_dict['introns in transcript'].append(len(splice_sites))
column_dict['strand'].append(strand)
chrom = df[df['transcript'] == tx].iloc[0]['chromosome']
column_dict['chromosome'].append(chrom)
column_dict['transcript size'].append((tx_dict[iso][1]-tx_dict[iso][0])/1000.)
# Check if first or last intron and add exon size
if n == 0:
column_dict['intron position'].append('First')
if strand == '+':
column_dict['exon size (us)'].append((five_site-tx_dict[iso][0])/1000.)
if len(splice_sites) > 1:
ds_length = (splice_sites[n+1][0] - three_site)/1000.
try:
if ds_length < 0:
ds_length = (splice_sites[n+2][0] - three_site)/1000.
except IndexError:
ds_length = np.NaN
else:
ds_length = (tx_dict[iso][1] - three_site)/1000.
elif strand == '-':
column_dict['exon size (us)'].append((tx_dict[iso][1]-five_site)/1000.)
if len(splice_sites) > 1:
ds_length = (three_site - splice_sites[n+1][0])/1000.
try:
if ds_length < 0:
ds_length = (three_site - splice_sites[n+2][0])/1000.
except IndexError:
ds_length = np.NaN
else:
ds_length = (three_site - tx_dict[iso][0])/1000.
column_dict['exon size (ds)'].append(ds_length)
elif n == len(splice_sites)-1:
column_dict['intron position'].append('Last')
column_dict['exon size (us)'].append((abs(five_site-splice_sites[n-1][1])-1)/1000.)
if strand == '+':
column_dict['exon size (ds)'].append((tx_dict[iso][1]-three_site)/1000.)
elif strand == '-':
column_dict['exon size (ds)'].append((three_site - tx_dict[iso][0])/1000.)
else:
column_dict['intron position'].append('Middle')
column_dict['exon size (us)'].append((abs(five_site-splice_sites[n-1][1])-1)/1000.)
column_dict['exon size (ds)'].append(abs(three_site - splice_sites[n+1][0])/1000.)
if in_df is True:
peak_index = chrom+':'+str(int(df_pos))
new_index.append(peak_index)
column_dict['position'].append(df_pos)
column_dict['3p score'].append(df.loc[peak_index,'3p score'])
column_dict['5p score'].append(df.loc[peak_index,'5p score'])
column_dict['alt splicing'].append(df.loc[peak_index,'alt splicing'])
column_dict['type'].append(df.loc[peak_index,'type'])
column_dict['peak'].append(True)
column_dict['seq5'].append(df.loc[peak_index,'seq5'])
column_dict['seq3'].append(df.loc[peak_index,'seq3'])
if in_df is False:
column_dict['alt splicing'].append(False)
column_dict['type'].append('5prime')
column_dict['peak'].append(False)
# Get position, index and sequence for scoring and position code
if strand == '+':
column_dict['position'].append(five_site+1)
new_index.append(chrom+':'+str(five_site+1))
sequence1 = fa_dict[chrom][(five_site-1):(five_site+7)]
sequence2 = fa_dict[chrom][(three_site-5):(three_site+3)]
elif strand == '-':
column_dict['position'].append(five_site-1)
new_index.append(chrom+':'+str(five_site-1))
sequence1 = fa_dict[chrom][(five_site-6):(five_site+2)]
sequence1 = SP.reverse_complement(sequence1)
sequence2 = fa_dict[chrom][(three_site-2):(three_site+6)]
sequence2 = SP.reverse_complement(sequence2)
column_dict['seq5'].append(sequence1)
column_dict['seq3'].append(sequence2)
# Score sequences
score_5, score_3 = SP.simple_score_junction(sequence1, sequence2, PSSM)
column_dict['3p score'].append(score_3)
column_dict['5p score'].append(score_5)
# Create new dataframe from column dictionary
new_df = pd.DataFrame(columns=column_dict.keys(), index=new_index)
for column, data in column_dict.iteritems():
new_df[column] = data
return new_df
def find_score_branches_ppy(quant_df, peak_branch_df, fa_dict):
#branches = generate_all_branches()
PSSM = branch_PSSM(peak_branch_df, fa_dict)
branches = []
if type(peak_branch_df) is not str:
for ix, branch in peak_branch_df['Branch seq'].iteritems():
seq = branch[2:7]
if seq[:-2] != 'A' and 'A' in seq:
A_ix = branch.rfind('A')
new_seq = branch[A_ix-3:A_ix+2]
if len(new_seq) == 5:
seq = new_seq
branches = peak_branch_df['Branch seq'].str[2:7]
else:
with open(peak_branch_df) as f:
for line in f:
branches.append(line.strip())
# Sort branches by abundance so that the most common ones are first in the search
br_abund = []
for branch in set(branches):
count = len([x for x in branches if x == branch])
if count > 1:
br_abund.append((branch, count))
br_abund = sorted(br_abund, key=lambda x: x[1], reverse=True)
branches = zip(*br_abund)[0]
branch_dict = collections.OrderedDict()
for branch in branches:
branch_dict[branch] = score_branch(branch, PSSM)
branch_3_dist = []
branch_score = []
branch_seqs = []
perc_py = []
for ix, r in quant_df.iterrows():
if r['strand'] == '+':
intron_seq = fa_dict[r['chromosome']][int(r['position']):int(r['position']+r['intron size'])]
three_site = r['position']+r['intron size']
elif r['strand'] == '-':
intron_seq = fa_dict[r['chromosome']][int(r['position']-r['intron size']-1):int(r['position']-1)]
intron_seq = SP.reverse_complement(intron_seq)
three_site = r['position']-r['intron size']
if type(peak_branch_df) is not str:
if ix in peak_branch_df['genome coord']:
ix_df = peak_branch_df[peak_branch_df['genome coord'] == ix]
ix_df = ix_df.sort_values('depth', ascending=False)
best_branch = ix_df.iloc[0,'branch site']
best_branch = abs(ix_df.iloc[0,'5p splice site']-best_branch)
seq = ix_df.iloc[0,'Branch seq'][2:7]
branch_seqs.append(seq)
branch_score.append(score_branch(seq, PSSM))
branch_3_dist.append(ix_df.iloc[0,'Branch to 3p distance'])
if 'N' in intron_seq[best_branch[0]+5:]:
print ix
print intron_seq
perc_py.append(percent_py(intron_seq[best_branch[0]+5:]))
else:
matches = []
for branch in branch_dict:
if branch in intron_seq:
matches.append((intron_seq.index(branch), branch, branch_dict[branch]))
if len(matches) == 0:
# Find the closest A
best_ix = intron_seq[:-3].rfind('A')
seq = intron_seq[best_ix-3:best_ix+2]
score = score_branch(seq, PSSM)
best_branch = (best_ix, seq, score)
#branch_3_dist.append(np.NaN)
#branch_score.append(np.NaN)
#branch_seqs.append('NNNNN')
#perc_py.append(percent_py(intron_seq[-30:]))
elif len(matches) > 1:
matches = sorted(matches, key=lambda x: x[2], reverse=True)
best_branch = matches[0]
else:
best_branch = matches[0]
branch_3_dist.append((len(intron_seq)-best_branch[0]-4)/1000.)
branch_score.append(best_branch[2])
branch_seqs.append(best_branch[1])
if len(intron_seq)-best_branch[0]-5 > 1:
if 'N' in intron_seq[best_branch[0]+5:]:
print ix
print intron_seq
perc_py.append(percent_py(intron_seq[best_branch[0]+5:]))
else:
perc_py.append(np.NaN)
else:
matches = []
for branch in branch_dict:
if branch in intron_seq:
matches.append((intron_seq.index(branch), branch, branch_dict[branch]))
if len(matches) == 0:
# Find the closest A
best_ix = intron_seq[:-3].rfind('A')
seq = intron_seq[best_ix-3:best_ix+2]
score = score_branch(seq, PSSM)
best_branch = (best_ix, seq, score)
#branch_3_dist.append(np.NaN)
#branch_score.append(np.NaN)
#branch_seqs.append('NNNNN')
#perc_py.append(percent_py(intron_seq[-30:]))
elif len(matches) > 1:
matches = sorted(matches, key=lambda x: x[2], reverse=True)
best_branch = matches[0]
else:
best_branch = matches[0]
branch_3_dist.append((len(intron_seq)-best_branch[0]-4)/1000.)
branch_score.append(best_branch[2])
branch_seqs.append(best_branch[1])
if len(intron_seq)-best_branch[0]-5 > 1:
if 'N' in intron_seq[best_branch[0]+5:]:
print ix
print intron_seq
perc_py.append(percent_py(intron_seq[best_branch[0]+5:]))
else:
perc_py.append(np.NaN)
print len(quant_df)
print len(branch_score)
quant_df['branch score'] = branch_score
quant_df['branch to 3p distance'] = branch_3_dist
quant_df['percent pPy'] = perc_py
branch_seqs = ['NNNNN' if len(x) < 5 else x for x in branch_seqs ]
print len(branch_seqs)
print len(quant_df)
for n in range(len(branch_seqs[0])):
pos = [x[n] for x in branch_seqs]
quant_df['branch-'+str(n)] = pos
print str(len(quant_df)-len(quant_df['branch score'].dropna()))+' introns without identifiable branches'
return quant_df
def build_junction_df(junction_bed, gff3_file, fasta, organism=None):
transcript_dict = SP.build_transcript_dict(gff3_file, organism=organism)
if type(fasta) == str:
fasta = SP.make_fasta_dict(fasta)
junction_dict = build_junction_dict(junction_bed,
gff3_file,
transcript_dict,
organism=organism)
junction_count = 0
for tx, junctions in junction_dict.iteritems():
junction_count += len(junctions)
junction_df = pd.DataFrame(index=range(junction_count),
columns=[
'intron tuple', 'chromosome', 'start',
'end', 'strand', 'depth', 'type', 'size',
'annotated intron size',
'annotated intron start',
'annotated intron end'
])
n = 0
for tx, junctions in junction_dict.iteritems():
for junction in junctions:
junction_df.ix[n] = [tx] + junction
n += 1
sequence1 = []
sequence2 = []
ann_seq1 = []
ann_seq2 = []
seq_type1 = []
seq_type2 = []
df_tx = []
for index, row in junction_df.iterrows():
df_tx.append(row['intron tuple'][0])
chrom = convert_chrom(row['chromosome'])
if row['strand'] == '+':
curr1 = fasta[chrom][(row['start'] - 1):(row['start'] + 7)]
sequence1.append(curr1)
curr2 = fasta[chrom][(row['end'] - 5):(row['end'] + 3)]
sequence2.append(curr2)
if row['annotated intron start'] is None:
ann_seq1.append(None)
ann_seq2.append(None)
else:
ann_seq1.append(
fasta[chrom][(row['annotated intron start'] -
1):(row['annotated intron start'] + 7)])
ann_seq2.append(fasta[chrom][(row['annotated intron end'] -
5):(row['annotated intron end'] +
3)])
elif row['strand'] == '-':
curr1 = SP.reverse_complement(fasta[chrom][(row['start'] -
6):(row['start'] + 2)])
sequence1.append(curr1)
curr2 = SP.reverse_complement(fasta[chrom][(row['end'] -
2):(row['end'] + 6)])
sequence2.append(curr2)
if row['annotated intron start'] is None:
ann_seq1.append(None)
ann_seq2.append(None)
else:
ann_seq1.append(
SP.reverse_complement(
fasta[chrom][row['annotated intron start'] -
6:row['annotated intron start'] + 2]))
ann_seq2.append(
SP.reverse_complement(
fasta[chrom][row['annotated intron end'] -
2:row['annotated intron end'] + 6]))
else:
sequence1.append('NNNNNNNN')
sequence2.append('NNNNNNNN')
ann_seq1.append('NNNNNNNN')
ann_seq2.append('NNNNNNNN')
if row['type'] == 'Annotated':
seq_type1.append('5p annotated')
seq_type2.append('3p annotated')
elif row['type'] == '5p tethered':
seq_type1.append('5p annotated')
seq_type2.append(curr2[4:6])
else:
seq_type1.append(curr1[2:4])
seq_type2.append(curr2[4:6])
junc_seq_df = junction_df
junc_seq_df['sequence1'] = sequence1
junc_seq_df['sequence2'] = sequence2
junc_seq_df['seq type1'] = seq_type1
junc_seq_df['seq type2'] = seq_type2
junc_seq_df['annotated sequence1'] = ann_seq1
junc_seq_df['annotated sequence2'] = ann_seq2
junc_seq_df['transcript'] = df_tx
return junc_seq_df
def generate_consensus_matrix(gff3, fasta_dict, PSSM=False):
#Populate gene dictionary and build genome
if 'pombe' in gff3.lower():
transcript_dict = SP.build_transcript_dict(gff3, organism='pombe')
ss, flag = SP.list_splice_sites(gff3, organism='pombe')
organism = 'pombe'
else:
transcript_dict = SP.build_transcript_dict(gff3)
ss, flag = SP.list_splice_sites(gff3)
organism = None
ss_dict = SP.collapse_ss_dict(ss)
genome = fasta_dict
#print genome.keys()
nuc_prob = gc_content(fasta_dict)
#print nuc_prob
base_dict = {"A":0, "C":1, "T":2, "G":3}
#First generate a consensus matrix for 5' and 3' splice site, where 1st row is A counts, second row is C, third row is T, fourth row is G.
pos_matrix_5prime = np.zeros([4,8])
pos_matrix_3prime = np.zeros([4,8])
counter1 = 0
counter2 = 0
for transcript, introns in ss_dict.iteritems():
counter2 += 1
if organism == 'pombe':
isoform = transcript+'.1'
else:
isoform = transcript+'T0'
strand = transcript_dict[isoform][2]
chrom = transcript_dict[isoform][3]
for intron in introns:
counter1+=1
if strand == '+':
seq = fasta_dict[chrom][(intron[0]-1):(intron[0]+7)]
elif strand == '-':
seq = fasta_dict[chrom][(intron[0]-6):(intron[0]+2)]
seq = SP.reverse_complement(seq)
for a, base in enumerate(seq):
pos_matrix_5prime[base_dict[base],a] += 1
if strand == '+':
seq = fasta_dict[chrom][(intron[1]-5):(intron[1]+3)]
elif strand == '-':
seq = fasta_dict[chrom][(intron[1]-2):(intron[1]+6)]
seq = SP.reverse_complement(seq)
for b, base in enumerate(seq):
pos_matrix_3prime[base_dict[base],b] += 1
#print counter1
#print counter2
float_formatter = lambda x: "%.1f" % x
np.set_printoptions(formatter={'float_kind':float_formatter})
a = 0
while a < 4:
b = 0
while b < 8:
if PSSM is False:
pos_matrix_5prime[a,b] = (pos_matrix_5prime[a,b])/float(counter1)
pos_matrix_3prime[a,b] = (pos_matrix_3prime[a,b])/float(counter1)
if PSSM is True:
if pos_matrix_5prime[a,b] == 0: pos_matrix_5prime[a,b] += 1
if pos_matrix_3prime[a,b] == 0: pos_matrix_3prime[a,b] += 1
pos_matrix_5prime[a,b] = np.log2((pos_matrix_5prime[a,b]/float(counter1))/nuc_prob[a])
pos_matrix_3prime[a,b] = np.log2((pos_matrix_3prime[a,b]/float(counter1))/nuc_prob[a])
b += 1
a += 1
return (pos_matrix_5prime, pos_matrix_3prime)
def get_junction_sequence(df, gff3_file, fasta_file):
df = df.sort_values('chr', axis=0)
#transcript_dict[transcript] = [start, end, strand, chromosome, CDS start, CDS end]
transcript_dict = SP.build_transcript_dict(gff3_file)
#splice_dict[transcipt] = [[5'sites][3'sites]]
splice_dict, flag = SP.list_splice_sites(gff3_file)
#fasta_dict[chr] = sequence
if type(fasta_file) is str:
fasta_dict = make_fasta_dict(fasta_file)
else:
fasta_dict = fasta_file
transcript_by_chr = {}
for transcript, coords in transcript_dict.iteritems():
chromosome = coords[3]
if chromosome in transcript_by_chr:
transcript_by_chr[chromosome].append(transcript)
else:
transcript_by_chr[chromosome] = []
transcript_by_chr[chromosome].append(transcript)
df['Gene'] = "Unknown"
df['intron'] = "Middle"
df['sequence1'] = ''
df['sequence2'] = ''
df['intron sequence'] = 'No sequence here'
n = 0
for n in range(len(df)):
coord1 = int(df['coord_1'][n].strip())
coord2 = int(df['coord_2'][n].strip())
chrom = df['chr'][n].strip()
strand = df['strand'][n].strip()
transcripts = transcript_by_chr[chrom]
for transcript in transcripts:
tx_strand = transcript_dict[transcript][2]
start = transcript_dict[transcript][0]
stop = transcript_dict[transcript][1]
if strand == tx_strand and coord1 >= start and coord2 <= stop:
df.loc[n,'Gene'] = transcript
if strand == '+':
sequence1 = fasta_dict[chrom][(coord1-3):(coord1+5)]
sequence2 = fasta_dict[chrom][(coord2-6):(coord2+2)]
all_seq = fasta_dict[chrom][(coord1-1):coord2]
elif strand == '-':
sequence1 = fasta_dict[chrom][(coord2-6):(coord2+2)]
sequence1 = SP.reverse_complement(sequence1)
sequence2 = fasta_dict[chrom][(coord1-3):(coord1+5)]
sequence2 = SP.reverse_complement(sequence2)
all_seq = fasta_dict[chrom][(coord1-1):coord2]
all_seq = SP.reverse_complement(all_seq)
df.loc[n,'sequence1'] = sequence1
df.loc[n,'sequence2'] = sequence2
df.loc[n,'intron sequence'] = all_seq
for transcript in transcripts:
if transcript in df['Gene'].tolist():
tx_df = df[df['Gene'] == transcript]
s = tx_df['coord_1']
min_idx = s.idxmin()
first = int(s.min())
#print transcript_dict[transcript][2]
#print first
max_idx = s.idxmax()
last = int(s.max())
#print last
if first == last:
df.loc[min_idx,'intron'] = 'Only'
else:
if transcript_dict[transcript][2] == '+':
df.loc[min_idx,'intron'] = 'First'
df.loc[max_idx,'intron'] = 'Last'
elif transcript_dict[transcript][2] == '-':
df.loc[min_idx,'intron'] = 'Last'
df.loc[max_idx,'intron'] = 'First'
for index, coord_1 in s.iteritems():
if df['intron'][index] == 'Middle':
if coord_1 in range(first-10, first+10):
df_idx = s[s == coord_1].index[0]
if transcript_dict[transcript][2] == '+':
df.loc[df_idx, 'intron'] = 'First'
elif transcript_dict[transcript][2] == '-':
df.loc[df_idx, 'intron'] = 'Last'
elif coord_1 in range(last-10, last+10):
df_idx = s[s == coord_1].index[0]
if transcript_dict[transcript][2] == '+':
df.loc[df_idx, 'intron'] = 'Last'
elif transcript_dict[transcript][2] == '-':
df.loc[df_idx, 'intron'] = 'First'
df = df[df['contained in'] != '']
df = df.reset_index()
return df
def quant_from_peak_df(peak_df, gff3, fa_dict, organism=None):
count1 = 0
count2 = 0
pssm = SP.generate_consensus_matrix(gff3, fa_dict, PSSM=True)
ss_dict, flag = SP.list_splice_sites(gff3, organism=organism)
ss_dict = SP.collapse_ss_dict(ss_dict)
quant_df = peak_df[(peak_df['type'] != '3prime')
& (peak_df['looks like'] != 'AG')]
quant_df['genome coord'] = quant_df['chromosome'].str.cat(
quant_df['position'].values.astype(str), sep=':')
quant_df.index = quant_df['genome coord']
quant_df = quant_df.drop('index', axis=1)
column_dict = {
'intron size': [],
'alt splicing': [],
'5p score': [],
'3p score': [],
'seq5': [],
'seq3': []
}
new_index = []
seq5 = []
seq3 = []
for coord in quant_df.index:
coord_df = quant_df[quant_df.index == coord]
three_site = None
alt3 = False
if len(coord_df) > 0:
coord_df = coord_df.sort_values('height', ascending=False).ix[0]
introns = ss_dict[coord_df['transcript']]
if 'prime' in coord_df['type']:
peak_range = range(coord_df['position'] - 5,
coord_df['position'] + 5)
for intron in introns:
if intron[0] in peak_range:
five_site = intron[0]
three_site = intron[1]
break
if len(quant_df[(quant_df['transcript'] == coord_df['transcript'])
& (quant_df['type'] == 'AG')]) > 0:
alt3 = True
else:
if 'AG' in quant_df[quant_df['transcript'] ==
coord_df['transcript']]['type']:
five_site = coord_df['position']
three_df = quant_df[
(quant_df['transcript'] == coord_df['transcript'])
& (quant_df['type'] == 'AG')]
three_df = three_df.sort_values('height', ascending=False)
three_site = three_df.ix[0]['position']
if three_site is not None:
new_index.append(coord)
size = abs(three_site - five_site) / 1000.
column_dict['intron size'].append(size)
column_dict['alt splicing'].append(alt3)
if coord_df['strand'] == '+':
s5 = fa_dict[coord_df['chromosome']][five_site - 2:five_site +
6]
s3 = fa_dict[coord_df['chromosome']][three_site -
6:three_site + 2]
elif coord_df['strand'] == '-':
s5 = fa_dict[coord_df['chromosome']][five_site - 6:five_site +
2]
s5 = SP.reverse_complement(s5)
s3 = fa_dict[coord_df['chromosome']][three_site -
2:three_site + 6]
s3 = SP.reverse_complement(s3)
column_dict['seq5'].append(s5)
column_dict['seq3'].append(s3)
scores = SP.simple_score_junction(s5, s3, pssm)
column_dict['3p score'].append(scores[1])
column_dict['5p score'].append(scores[0])
new_quant_df = quant_df[quant_df.index.isin(new_index)][[
'genome coord', 'chromosome', 'strand', 'transcript', 'position',
'type'
]]
for column, data in column_dict.iteritems():
new_quant_df[column] = data
new_quant_df = new_quant_df.drop_duplicates(
subset='genome coord', keep='first').set_index('genome coord')
new_quant_df = SP.backfill_splice_sites(new_quant_df,
gff3,
fa_dict,
pssm,
organism=organism)
#for n in range(len(new_quant_df['seq5'].iloc[0])):
# new_quant_df['Base 5-'+str(n)] = [x[n] for x in new_quant_df['seq5']]
#for n in range(len(new_quant_df['seq3'].iloc[0])):
# new_quant_df['Base 3-'+str(n)] = [x[n] for x in new_quant_df['seq3']]
#new_quant_df = new_quant_df.drop(['seq5','seq3'], axis=1)
new_quant_df = SP.find_score_branches_ppy(
new_quant_df, '/home/jordan/GENOMES/S288C/S288C_branches2.txt',
fa_dict)
return new_quant_df
def make_quant_df(junc_df, branch_df, gff3, fa_dict, organism=None):
pssm = SP.generate_consensus_matrix(gff3, fa_dict, PSSM=True)
quant_df = junc_df[(junc_df['type'] != '3prime')
& (junc_df['looks like'] != 'AG')]
new_intron_size = []
alt_splice = []
score_3 = []
score_5 = []
seq5 = []
seq3 = []
new_quant_df = pd.DataFrame(index=set(quant_df.index),
columns=['intron size', 'alt splicing'])
for coord in new_quant_df.index:
coord_df = quant_df[quant_df.index == coord]
#Determine if multiple junctions come from this peak
if len(coord_df) > 1: alt_splice.append(True)
else: alt_splice.append(False)
if max(coord_df['annotated intron size']) > 0:
coord_df = coord_df.sort_values('annotated intron size',
ascending=False)
new_intron_size.append(coord_df.ix[0]['annotated intron size'] /
1000.)
seq5.append(coord_df.ix[0]['junction sequence1'])
seq3.append(coord_df.ix[0]['junction sequence2'])
score_3.append(coord_df.ix[0]['annotated 3p score'])
score_5.append(coord_df.ix[0]['annotated 5p score'])
else:
coord_df = coord_df.sort_values('junction size', ascending=False)
new_intron_size.append(coord_df.ix[0]['junction size'] / 1000.)
seq5.append(coord_df.ix[0]['junction sequence1'])
seq3.append(coord_df.ix[0]['junction sequence2'])
scores = SP.simple_score_junction(
coord_df.ix[0]['junction sequence1'],
coord_df.ix[0]['junction sequence2'], pssm)
score_3.append(scores[1])
score_5.append(scores[0])
new_quant_df['intron size'] = new_intron_size
new_quant_df['alt splicing'] = alt_splice
new_quant_df['5p score'] = score_5
new_quant_df['3p score'] = score_3
new_quant_df['seq5'] = seq5
new_quant_df['seq3'] = seq3
quant_df = quant_df.sort_values('annotated intron size')
quant_df = quant_df.reset_index(drop=True).drop_duplicates(
subset='genome coord', keep='first').set_index('genome coord')
new_quant_df = new_quant_df.merge(
quant_df[['transcript', 'chromosome', 'position', 'strand', 'type']],
right_index=True,
left_index=True)
for coord in set(branch_df['genome coord']):
if coord not in new_quant_df.index:
coord_df = branch_df[branch_df['genome coord'] == coord]
coord_df = coord_df.sort_values('depth')
best = coord_df.iloc[0]
coord_dict = {
'transcript': best['transcript'][:-2],
'chromosome': best['chromosome'],
'position': best['5p splice site'],
'strand': best['strand'],
'type': best['type'],
'intron size': best['intron size'],
'alt splicing': np.where(len(coord_df) > 1, True, False),
'5p score': np.NaN,
'3p score': np.NaN,
'seq5': '',
'seq3': ''
}
if len(best['5p seq']) > 0:
coord_dict['seq5'] = best['5p seq']
else:
if best['strand'] == '+':
coord_dict['seq5'] = fa_dict[best['chromosome']][(
int(best['5p splice site']) -
1):(int(best['5p splice site']) + 7)]
elif best['strand'] == '-':
coord_dict['seq5'] = fa_dict[best['chromosome']][(
int(best['5p splice site']) -
6):(int(best['5p splice site']) + 2)]
coord_dict['seq5'] = SP.reverse_complement(
coord_dict['seq5'])
if str(best['3p splice site']) != 'nan':
three_site = best['3p splice site']
else:
if best['strand'] == '+':
after_branch = fa_dict[best['chromosome']][
best['branch site']:best['branch site'] + 100]
elif best['strand'] == '-':
after_branch = fa_dict[
best['chromosome']][best['branch site'] -
100:best['branch site']]
after_branch = SP.reverse_complement(after_branch)
for subs in ['TAG', 'CAG', 'GAG', 'AAG']:
if subs in after_branch:
ix = after_branch.find(subs) + 3
break
three_site = best['branch site'] + ix
if best['strand'] == '-':
three_site = best['branch site'] - ix
coord_dict['intron size'] = abs(coord_dict['position'] -
three_site)
if best['strand'] == '+':
coord_dict['seq3'] = fa_dict[
best['chromosome']][int(three_site - 5):int(three_site) +
3]
elif best['strand'] == '-':
coord_dict['seq3'] = fa_dict[
best['chromosome']][int(three_site) - 2:int(three_site) +
6]
coord_dict['seq3'] = SP.reverse_complement(coord_dict['seq3'])
coord_dict['5p score'], coord_dict[
'3p score'] = SP.simple_score_junction(coord_dict['seq5'],
coord_dict['seq3'],
pssm)
coord_s = pd.Series(coord_dict, name=coord)
new_quant_df = new_quant_df.append(coord_s)
new_quant_df = backfill_splice_sites(new_quant_df,
gff3,
fa_dict,
pssm,
organism=organism)
for n in range(len(new_quant_df['seq5'].iloc[0])):
new_quant_df['Base 5-' + str(n)] = [x[n] for x in new_quant_df['seq5']]
for n in range(len(new_quant_df['seq3'].iloc[0])):
new_quant_df['Base 3-' + str(n)] = [x[n] for x in new_quant_df['seq3']]
new_quant_df = new_quant_df.drop(['seq5', 'seq3'], axis=1)
lariat_df = junc_df[(junc_df['type'] == '3prime') |
(junc_df['looks like'] == 'AG')]
lariat_df = lariat_df.sort_values(
['genome coord', 'annotated intron size'], ascending=False)
lariat_df = lariat_df.reset_index(drop=True).drop_duplicates(
subset='genome coord', keep='first').set_index('genome coord')
lariat_df = lariat_df[[
'transcript', 'chromosome', 'position', 'strand', 'type'
]]
return new_quant_df, lariat_df
def backfill_splice_sites(df, gff3, fa_dict, PSSM, organism=None):
tx_dict = SP.build_transcript_dict(gff3, organism=organism)
ss_dict, flag = SP.list_splice_sites(gff3, organism=organism)
ss_dict = SP.collapse_ss_dict(ss_dict)
column_dict = {
'position': [],
'transcript': [],
'alt splicing': [],
'type': [],
'strand': [],
'introns in transcript': [],
'intron size': [],
'chromosome': [],
'5p score': [],
'3p score': [],
'intron position': [],
'exon size (us)': [],
'exon size (ds)': [],
'transcript size': [],
'peak': [],
'seq5': [],
'seq3': []
}
new_index = []
for tx in set(df['transcript']):
strand = df[df['transcript'] == tx].iloc[0]['strand']
splice_sites = ss_dict[tx]
if strand == '+':
splice_sites = sorted(list(splice_sites), key=lambda x: x[0])
elif strand == '-':
splice_sites = sorted(list(splice_sites),
key=lambda x: x[0],
reverse=True)
df_pos = None
for n, (five_site, three_site) in enumerate(splice_sites):
# Check if already in dataframe
in_df = False
for peak in df[df['transcript'] == tx]['position']:
if five_site in range(int(peak) - 5, int(peak) + 5):
in_df = True
df_pos = peak
break
column_dict['transcript'].append(tx)
if organism == 'pombe':
iso = tx + '.1'
else:
iso = tx + 'T0'
column_dict['intron size'].append(abs(three_site - five_site))
column_dict['introns in transcript'].append(len(splice_sites))
column_dict['strand'].append(strand)
chrom = df[df['transcript'] == tx].iloc[0]['chromosome']
column_dict['chromosome'].append(chrom)
column_dict['transcript size'].append(
(tx_dict[iso][1] - tx_dict[iso][0]) / 1000.)
# Check if first or last intron and add exon size
if n == 0:
column_dict['intron position'].append('First')
if strand == '+':
column_dict['exon size (us)'].append(
(five_site - tx_dict[iso][0]) / 1000.)
if len(splice_sites) > 1:
ds_length = (splice_sites[n + 1][0] -
three_site) / 1000.
try:
if ds_length < 0:
ds_length = (splice_sites[n + 2][0] -
three_site) / 1000.
except IndexError:
ds_length = np.NaN
else:
ds_length = (tx_dict[iso][1] - three_site) / 1000.
elif strand == '-':
column_dict['exon size (us)'].append(
(tx_dict[iso][1] - five_site) / 1000.)
if len(splice_sites) > 1:
ds_length = (three_site -
splice_sites[n + 1][0]) / 1000.
try:
if ds_length < 0:
ds_length = (three_site -
splice_sites[n + 2][0]) / 1000.
except IndexError:
ds_length = np.NaN
else:
ds_length = (three_site - tx_dict[iso][0]) / 1000.
column_dict['exon size (ds)'].append(ds_length)
elif n == len(splice_sites) - 1:
column_dict['intron position'].append('Last')
column_dict['exon size (us)'].append(
(abs(five_site - splice_sites[n - 1][1]) - 1) / 1000.)
if strand == '+':
column_dict['exon size (ds)'].append(
(tx_dict[iso][1] - three_site) / 1000.)
elif strand == '-':
column_dict['exon size (ds)'].append(
(three_site - tx_dict[iso][0]) / 1000.)
else:
column_dict['intron position'].append('Middle')
column_dict['exon size (us)'].append(
(abs(five_site - splice_sites[n - 1][1]) - 1) / 1000.)
column_dict['exon size (ds)'].append(
abs(three_site - splice_sites[n + 1][0]) / 1000.)
if in_df is True:
peak_index = chrom + ':' + str(int(df_pos))
new_index.append(peak_index)
column_dict['position'].append(df_pos)
column_dict['3p score'].append(df.loc[peak_index, '3p score'])
column_dict['5p score'].append(df.loc[peak_index, '5p score'])
column_dict['alt splicing'].append(df.loc[peak_index,
'alt splicing'])
column_dict['type'].append(df.loc[peak_index, 'type'])
column_dict['peak'].append(True)
column_dict['seq5'].append(df.loc[peak_index, 'seq5'])
column_dict['seq3'].append(df.loc[peak_index, 'seq3'])
if in_df is False:
column_dict['alt splicing'].append(False)
column_dict['type'].append('5prime')
column_dict['peak'].append(False)
# Get position, index and sequence for scoring and position code
if strand == '+':
column_dict['position'].append(five_site + 1)
new_index.append(chrom + ':' + str(five_site + 1))
sequence1 = fa_dict[chrom][(five_site - 1):(five_site + 7)]
sequence2 = fa_dict[chrom][(three_site - 5):(three_site +
3)]
elif strand == '-':
column_dict['position'].append(five_site - 1)
new_index.append(chrom + ':' + str(five_site - 1))
sequence1 = fa_dict[chrom][(five_site - 6):(five_site + 2)]
sequence1 = SP.reverse_complement(sequence1)
sequence2 = fa_dict[chrom][(three_site - 2):(three_site +
6)]
sequence2 = SP.reverse_complement(sequence2)
column_dict['seq5'].append(sequence1)
column_dict['seq3'].append(sequence2)
# Score sequences
score_5, score_3 = SP.simple_score_junction(
sequence1, sequence2, PSSM)
column_dict['3p score'].append(score_3)
column_dict['5p score'].append(score_5)
# Create new dataframe from column dictionary
new_df = pd.DataFrame(columns=column_dict.keys(), index=new_index)
for column, data in column_dict.iteritems():
new_df[column] = data
return new_df
def find_score_branches_ppy(quant_df, peak_branch_df, fa_dict):
#branches = generate_all_branches()
PSSM = branch_PSSM(peak_branch_df, fa_dict)
branches = []
if type(peak_branch_df) is not str:
for ix, branch in peak_branch_df['Branch seq'].iteritems():
seq = branch[2:7]
if seq[:-2] != 'A' and 'A' in seq:
A_ix = branch.rfind('A')
new_seq = branch[A_ix - 3:A_ix + 2]
if len(new_seq) == 5:
seq = new_seq
branches = peak_branch_df['Branch seq'].str[2:7]
else:
with open(peak_branch_df) as f:
for line in f:
branches.append(line.strip())
# Sort branches by abundance so that the most common ones are first in the search
br_abund = []
for branch in set(branches):
count = len([x for x in branches if x == branch])
if count > 1:
br_abund.append((branch, count))
br_abund = sorted(br_abund, key=lambda x: x[1], reverse=True)
branches = zip(*br_abund)[0]
branch_dict = collections.OrderedDict()
for branch in branches:
branch_dict[branch] = score_branch(branch, PSSM)
branch_3_dist = []
branch_score = []
branch_seqs = []
perc_py = []
for ix, r in quant_df.iterrows():
if r['strand'] == '+':
intron_seq = fa_dict[
r['chromosome']][int(r['position']):int(r['position'] +
r['intron size'])]
three_site = r['position'] + r['intron size']
elif r['strand'] == '-':
intron_seq = fa_dict[r['chromosome']][int(r['position'] -
r['intron size'] -
1):int(r['position'] -
1)]
intron_seq = SP.reverse_complement(intron_seq)
three_site = r['position'] - r['intron size']
if type(peak_branch_df) is not str:
if ix in peak_branch_df['genome coord']:
ix_df = peak_branch_df[peak_branch_df['genome coord'] == ix]
ix_df = ix_df.sort_values('depth', ascending=False)
best_branch = ix_df.iloc[0, 'branch site']
best_branch = abs(ix_df.iloc[0, '5p splice site'] -
best_branch)
seq = ix_df.iloc[0, 'Branch seq'][2:7]
branch_seqs.append(seq)
branch_score.append(score_branch(seq, PSSM))
branch_3_dist.append(ix_df.iloc[0, 'Branch to 3p distance'])
if 'N' in intron_seq[best_branch[0] + 5:]:
print ix
print intron_seq
perc_py.append(percent_py(intron_seq[best_branch[0] + 5:]))
else:
matches = []
for branch in branch_dict:
if branch in intron_seq:
matches.append((intron_seq.index(branch), branch,
branch_dict[branch]))
if len(matches) == 0:
# Find the closest A
best_ix = intron_seq[:-3].rfind('A')
seq = intron_seq[best_ix - 3:best_ix + 2]
score = score_branch(seq, PSSM)
best_branch = (best_ix, seq, score)
#branch_3_dist.append(np.NaN)
#branch_score.append(np.NaN)
#branch_seqs.append('NNNNN')
#perc_py.append(percent_py(intron_seq[-30:]))
elif len(matches) > 1:
matches = sorted(matches, key=lambda x: x[2], reverse=True)
best_branch = matches[0]
else:
best_branch = matches[0]
branch_3_dist.append(
(len(intron_seq) - best_branch[0] - 4) / 1000.)
branch_score.append(best_branch[2])
branch_seqs.append(best_branch[1])
if len(intron_seq) - best_branch[0] - 5 > 1:
if 'N' in intron_seq[best_branch[0] + 5:]:
print ix
print intron_seq
perc_py.append(percent_py(intron_seq[best_branch[0] + 5:]))
else:
perc_py.append(np.NaN)
else:
matches = []
for branch in branch_dict:
if branch in intron_seq:
matches.append((intron_seq.index(branch), branch,
branch_dict[branch]))
if len(matches) == 0:
# Find the closest A
best_ix = intron_seq[:-3].rfind('A')
seq = intron_seq[best_ix - 3:best_ix + 2]
score = score_branch(seq, PSSM)
best_branch = (best_ix, seq, score)
#branch_3_dist.append(np.NaN)
#branch_score.append(np.NaN)
#branch_seqs.append('NNNNN')
#perc_py.append(percent_py(intron_seq[-30:]))
elif len(matches) > 1:
matches = sorted(matches, key=lambda x: x[2], reverse=True)
best_branch = matches[0]
else:
best_branch = matches[0]
branch_3_dist.append(
(len(intron_seq) - best_branch[0] - 4) / 1000.)
branch_score.append(best_branch[2])
branch_seqs.append(best_branch[1])
if len(intron_seq) - best_branch[0] - 5 > 1:
if 'N' in intron_seq[best_branch[0] + 5:]:
print ix
print intron_seq
perc_py.append(percent_py(intron_seq[best_branch[0] + 5:]))
else:
perc_py.append(np.NaN)
print len(quant_df)
print len(branch_score)
quant_df['branch score'] = branch_score
quant_df['branch to 3p distance'] = branch_3_dist
quant_df['percent pPy'] = perc_py
branch_seqs = ['NNNNN' if len(x) < 5 else x for x in branch_seqs]
print len(branch_seqs)
print len(quant_df)
for n in range(len(branch_seqs[0])):
pos = [x[n] for x in branch_seqs]
quant_df['branch-' + str(n)] = pos
print str(len(quant_df) - len(quant_df['branch score'].dropna())
) + ' introns without identifiable branches'
return quant_df
def collect_intron_seq(gff3_file,
fasta_file,
ss_dict=None,
junction_bed=None,
gene_list=None,
peak_df=None,
organism=None):
transcript_dict = SP.build_transcript_dict(gff3_file, organism=organism)
if type(fasta_file) == dict:
fasta_dict = fasta_file
elif fasta_file.endswith('json'):
with open(fasta_file, 'r') as f:
fasta_dict = json.load(f)
else:
fasta_dict = make_fasta_dict(fasta_file)
if ss_dict is not None:
ss_dict = ss_dict
elif junction_bed is not None:
ss_dict = SP.build_junction_dict(junction_bed,
gff3_file,
transcript_dict,
organism=organism)
elif peak_df is not None:
ss_dict = {}
peak_df = peak_df[~peak_df['type'].str.contains('prime')]
for ix, r in peak_df.iterrows():
if r['transcript'] not in ss_dict:
ss_dict[r['transcript']] = []
if r['strand'] == '+':
ss_dict[r['transcript']].append(
(r['position'], r['position'] + 50))
elif r['strand'] == '-':
ss_dict[r['transcript']].append(
(r['position'], r['position'] - 50))
else:
ss_dict, intron_flag = SP.list_splice_sites(gff3_file,
gene_list=gene_list,
organism=organism)
ss_dict = SP.collapse_ss_dict(ss_dict)
seq_dict = {}
for transcript, introns in ss_dict.iteritems():
if junction_bed is None:
if organism == 'pombe':
transcript = transcript + '.1'
else:
transcript = transcript + 'T0'
introns = list(introns)
strand = transcript_dict[transcript][2]
chrom = transcript_dict[transcript][3]
n = 0
for n in range(len(introns)):
if strand == '+':
seq_dict[transcript + '-' + chrom + ':' +
str(introns[n][0] +
1)] = fasta_dict[chrom][introns[n][0] +
2:introns[n][0] + 17]
elif strand == '-':
seq = fasta_dict[chrom][introns[n][0] - 16:introns[n][0] - 1]
seq_dict[transcript + '-' + chrom + ':' +
str(introns[n][0])] = SP.reverse_complement(seq)
return seq_dict
def make_quant_df(junc_df, branch_df, gff3, fa_dict, organism=None):
pssm = SP.generate_consensus_matrix(gff3, fa_dict, PSSM=True)
quant_df = junc_df[(junc_df['type'] != '3prime') & (junc_df['looks like'] != 'AG')]
new_intron_size = []
alt_splice = []
score_3 = []
score_5 = []
seq5 = []
seq3 = []
new_quant_df = pd.DataFrame(index=set(quant_df.index), columns=['intron size','alt splicing'])
for coord in new_quant_df.index:
coord_df = quant_df[quant_df.index == coord]
#Determine if multiple junctions come from this peak
if len(coord_df) > 1: alt_splice.append(True)
else: alt_splice.append(False)
if max(coord_df['annotated intron size']) > 0:
coord_df = coord_df.sort_values('annotated intron size', ascending=False)
new_intron_size.append(coord_df.ix[0]['annotated intron size']/1000.)
seq5.append(coord_df.ix[0]['junction sequence1'])
seq3.append(coord_df.ix[0]['junction sequence2'])
score_3.append(coord_df.ix[0]['annotated 3p score'])
score_5.append(coord_df.ix[0]['annotated 5p score'])
else:
coord_df = coord_df.sort_values('junction size', ascending=False)
new_intron_size.append(coord_df.ix[0]['junction size']/1000.)
seq5.append(coord_df.ix[0]['junction sequence1'])
seq3.append(coord_df.ix[0]['junction sequence2'])
scores = SP.simple_score_junction(coord_df.ix[0]['junction sequence1'], coord_df.ix[0]['junction sequence2'], pssm)
score_3.append(scores[1])
score_5.append(scores[0])
new_quant_df['intron size'] = new_intron_size
new_quant_df['alt splicing'] = alt_splice
new_quant_df['5p score'] = score_5
new_quant_df['3p score'] = score_3
new_quant_df['seq5'] = seq5
new_quant_df['seq3'] = seq3
quant_df = quant_df.sort_values('annotated intron size')
quant_df = quant_df.reset_index(drop=True).drop_duplicates(subset='genome coord', keep='first').set_index('genome coord')
new_quant_df = new_quant_df.merge(quant_df[['transcript','chromosome','position','strand','type']], right_index=True, left_index=True)
for coord in set(branch_df['genome coord']):
if coord not in new_quant_df.index:
coord_df = branch_df[branch_df['genome coord'] == coord]
coord_df = coord_df.sort_values('depth')
best = coord_df.iloc[0]
coord_dict = {'transcript':best['transcript'][:-2],
'chromosome':best['chromosome'],
'position':best['5p splice site'],
'strand':best['strand'],
'type':best['type'],
'intron size':best['intron size'],
'alt splicing':np.where(len(coord_df)> 1, True, False),
'5p score':np.NaN,
'3p score':np.NaN,
'seq5':'','seq3':''}
if len(best['5p seq']) > 0:
coord_dict['seq5'] = best['5p seq']
else:
if best['strand'] == '+':
coord_dict['seq5'] = fa_dict[best['chromosome']][(int(best['5p splice site'])-1):(int(best['5p splice site'])+7)]
elif best['strand'] == '-':
coord_dict['seq5'] = fa_dict[best['chromosome']][(int(best['5p splice site'])-6):(int(best['5p splice site'])+2)]
coord_dict['seq5'] = SP.reverse_complement(coord_dict['seq5'])
if str(best['3p splice site']) != 'nan':
three_site = best['3p splice site']
else:
if best['strand'] == '+':
after_branch = fa_dict[best['chromosome']][best['branch site']:best['branch site']+100]
elif best['strand'] == '-':
after_branch = fa_dict[best['chromosome']][best['branch site']-100:best['branch site']]
after_branch = SP.reverse_complement(after_branch)
for subs in ['TAG','CAG','GAG','AAG']:
if subs in after_branch:
ix = after_branch.find(subs)+3
break
three_site = best['branch site']+ix
if best['strand'] == '-':
three_site = best['branch site']-ix
coord_dict['intron size'] = abs(coord_dict['position']-three_site)
if best['strand'] == '+':
coord_dict['seq3'] = fa_dict[best['chromosome']][int(three_site-5):int(three_site)+3]
elif best['strand'] == '-':
coord_dict['seq3'] = fa_dict[best['chromosome']][int(three_site)-2:int(three_site)+6]
coord_dict['seq3'] = SP.reverse_complement(coord_dict['seq3'])
coord_dict['5p score'], coord_dict['3p score'] = SP.simple_score_junction(coord_dict['seq5'], coord_dict['seq3'], pssm)
coord_s = pd.Series(coord_dict, name=coord)
new_quant_df = new_quant_df.append(coord_s)
new_quant_df = backfill_splice_sites(new_quant_df, gff3, fa_dict, pssm, organism=organism)
for n in range(len(new_quant_df['seq5'].iloc[0])):
new_quant_df['Base 5-'+str(n)] = [x[n] for x in new_quant_df['seq5']]
for n in range(len(new_quant_df['seq3'].iloc[0])):
new_quant_df['Base 3-'+str(n)] = [x[n] for x in new_quant_df['seq3']]
new_quant_df = new_quant_df.drop(['seq5','seq3'], axis=1)
lariat_df = junc_df[(junc_df['type'] == '3prime') | (junc_df['looks like'] == 'AG')]
lariat_df = lariat_df.sort_values(['genome coord','annotated intron size'], ascending=False)
lariat_df = lariat_df.reset_index(drop=True).drop_duplicates(subset='genome coord', keep='first').set_index('genome coord')
lariat_df = lariat_df[['transcript','chromosome','position','strand','type']]
return new_quant_df, lariat_df
def generate_consensus_matrix(gff3, fasta_dict, PSSM=False):
#Populate gene dictionary and build genome
if 'pombe' in gff3.lower():
transcript_dict = SP.build_transcript_dict(gff3, organism='pombe')
ss, flag = SP.list_splice_sites(gff3, organism='pombe')
organism = 'pombe'
else:
transcript_dict = SP.build_transcript_dict(gff3)
ss, flag = SP.list_splice_sites(gff3)
organism = None
ss_dict = SP.collapse_ss_dict(ss)
genome = fasta_dict
#print genome.keys()
nuc_prob = gc_content(fasta_dict)
#print nuc_prob
base_dict = {"A": 0, "C": 1, "T": 2, "G": 3}
#First generate a consensus matrix for 5' and 3' splice site, where 1st row is A counts, second row is C, third row is T, fourth row is G.
pos_matrix_5prime = np.zeros([4, 8])
pos_matrix_3prime = np.zeros([4, 8])
counter1 = 0
counter2 = 0
for transcript, introns in ss_dict.iteritems():
counter2 += 1
if organism == 'pombe':
isoform = transcript + '.1'
else:
isoform = transcript + 'T0'
strand = transcript_dict[isoform][2]
chrom = transcript_dict[isoform][3]
for intron in introns:
counter1 += 1
if strand == '+':
seq = fasta_dict[chrom][(intron[0] - 1):(intron[0] + 7)]
elif strand == '-':
seq = fasta_dict[chrom][(intron[0] - 6):(intron[0] + 2)]
seq = SP.reverse_complement(seq)
for a, base in enumerate(seq):
pos_matrix_5prime[base_dict[base], a] += 1
if strand == '+':
seq = fasta_dict[chrom][(intron[1] - 5):(intron[1] + 3)]
elif strand == '-':
seq = fasta_dict[chrom][(intron[1] - 2):(intron[1] + 6)]
seq = SP.reverse_complement(seq)
for b, base in enumerate(seq):
pos_matrix_3prime[base_dict[base], b] += 1
#print counter1
#print counter2
float_formatter = lambda x: "%.1f" % x
np.set_printoptions(formatter={'float_kind': float_formatter})
a = 0
while a < 4:
b = 0
while b < 8:
if PSSM is False:
pos_matrix_5prime[a,
b] = (pos_matrix_5prime[a,
b]) / float(counter1)
pos_matrix_3prime[a,
b] = (pos_matrix_3prime[a,
b]) / float(counter1)
if PSSM is True:
if pos_matrix_5prime[a, b] == 0: pos_matrix_5prime[a, b] += 1
if pos_matrix_3prime[a, b] == 0: pos_matrix_3prime[a, b] += 1
pos_matrix_5prime[a, b] = np.log2(
(pos_matrix_5prime[a, b] / float(counter1)) / nuc_prob[a])
pos_matrix_3prime[a, b] = np.log2(
(pos_matrix_3prime[a, b] / float(counter1)) / nuc_prob[a])
b += 1
a += 1
return (pos_matrix_5prime, pos_matrix_3prime)
def get_junction_sequence(df, gff3_file, fasta_file):
df = df.sort_values('chr', axis=0)
#transcript_dict[transcript] = [start, end, strand, chromosome, CDS start, CDS end]
transcript_dict = SP.build_transcript_dict(gff3_file)
#splice_dict[transcipt] = [[5'sites][3'sites]]
splice_dict, flag = SP.list_splice_sites(gff3_file)
#fasta_dict[chr] = sequence
if type(fasta_file) is str:
fasta_dict = make_fasta_dict(fasta_file)
else:
fasta_dict = fasta_file
transcript_by_chr = {}
for transcript, coords in transcript_dict.iteritems():
chromosome = coords[3]
if chromosome in transcript_by_chr:
transcript_by_chr[chromosome].append(transcript)
else:
transcript_by_chr[chromosome] = []
transcript_by_chr[chromosome].append(transcript)
df['Gene'] = "Unknown"
df['intron'] = "Middle"
df['sequence1'] = ''
df['sequence2'] = ''
df['intron sequence'] = 'No sequence here'
n = 0
for n in range(len(df)):
coord1 = int(df['coord_1'][n].strip())
coord2 = int(df['coord_2'][n].strip())
chrom = df['chr'][n].strip()
strand = df['strand'][n].strip()
transcripts = transcript_by_chr[chrom]
for transcript in transcripts:
tx_strand = transcript_dict[transcript][2]
start = transcript_dict[transcript][0]
stop = transcript_dict[transcript][1]
if strand == tx_strand and coord1 >= start and coord2 <= stop:
df.loc[n, 'Gene'] = transcript
if strand == '+':
sequence1 = fasta_dict[chrom][(coord1 - 3):(coord1 + 5)]
sequence2 = fasta_dict[chrom][(coord2 - 6):(coord2 + 2)]
all_seq = fasta_dict[chrom][(coord1 - 1):coord2]
elif strand == '-':
sequence1 = fasta_dict[chrom][(coord2 - 6):(coord2 + 2)]
sequence1 = SP.reverse_complement(sequence1)
sequence2 = fasta_dict[chrom][(coord1 - 3):(coord1 + 5)]
sequence2 = SP.reverse_complement(sequence2)
all_seq = fasta_dict[chrom][(coord1 - 1):coord2]
all_seq = SP.reverse_complement(all_seq)
df.loc[n, 'sequence1'] = sequence1
df.loc[n, 'sequence2'] = sequence2
df.loc[n, 'intron sequence'] = all_seq
for transcript in transcripts:
if transcript in df['Gene'].tolist():
tx_df = df[df['Gene'] == transcript]
s = tx_df['coord_1']
min_idx = s.idxmin()
first = int(s.min())
#print transcript_dict[transcript][2]
#print first
max_idx = s.idxmax()
last = int(s.max())
#print last
if first == last:
df.loc[min_idx, 'intron'] = 'Only'
else:
if transcript_dict[transcript][2] == '+':
df.loc[min_idx, 'intron'] = 'First'
df.loc[max_idx, 'intron'] = 'Last'
elif transcript_dict[transcript][2] == '-':
df.loc[min_idx, 'intron'] = 'Last'
df.loc[max_idx, 'intron'] = 'First'
for index, coord_1 in s.iteritems():
if df['intron'][index] == 'Middle':
if coord_1 in range(first - 10, first + 10):
df_idx = s[s == coord_1].index[0]
if transcript_dict[transcript][2] == '+':
df.loc[df_idx, 'intron'] = 'First'
elif transcript_dict[transcript][2] == '-':
df.loc[df_idx, 'intron'] = 'Last'
elif coord_1 in range(last - 10, last + 10):
df_idx = s[s == coord_1].index[0]
if transcript_dict[transcript][2] == '+':
df.loc[df_idx, 'intron'] = 'Last'
elif transcript_dict[transcript][2] == '-':
df.loc[df_idx, 'intron'] = 'First'
df = df[df['contained in'] != '']
df = df.reset_index()
return df
def quant_from_peak_df(peak_df, gff3, fa_dict, organism=None):
count1 = 0
count2 = 0
pssm = SP.generate_consensus_matrix(gff3, fa_dict, PSSM=True)
ss_dict, flag = SP.list_splice_sites(gff3, organism=organism)
ss_dict = SP.collapse_ss_dict(ss_dict)
quant_df = peak_df[(peak_df['type'] != '3prime') & (peak_df['looks like'] != 'AG')]
quant_df['genome coord'] = quant_df['chromosome'].str.cat(quant_df['position'].values.astype(str), sep=':')
quant_df.index = quant_df['genome coord']
quant_df = quant_df.drop('index', axis=1)
column_dict = {'intron size':[], 'alt splicing':[], '5p score':[], '3p score':[], 'seq5':[], 'seq3':[]}
new_index = []
seq5 = []
seq3 = []
for coord in quant_df.index:
coord_df = quant_df[quant_df.index == coord]
three_site = None
alt3 = False
if len(coord_df) > 0:
coord_df = coord_df.sort_values('height', ascending=False).ix[0]
introns = ss_dict[coord_df['transcript']]
if 'prime' in coord_df['type']:
peak_range = range(coord_df['position']-5,coord_df['position']+5)
for intron in introns:
if intron[0] in peak_range:
five_site = intron[0]
three_site = intron[1]
break
if len(quant_df[(quant_df['transcript'] == coord_df['transcript']) & (quant_df['type'] == 'AG')]) > 0:
alt3=True
else:
if 'AG' in quant_df[quant_df['transcript'] == coord_df['transcript']]['type']:
five_site = coord_df['position']
three_df = quant_df[(quant_df['transcript'] == coord_df['transcript']) & (quant_df['type'] == 'AG')]
three_df = three_df.sort_values('height', ascending=False)
three_site = three_df.ix[0]['position']
if three_site is not None:
new_index.append(coord)
size = abs(three_site-five_site)/1000.
column_dict['intron size'].append(size)
column_dict['alt splicing'].append(alt3)
if coord_df['strand'] == '+':
s5 = fa_dict[coord_df['chromosome']][five_site-2:five_site+6]
s3 = fa_dict[coord_df['chromosome']][three_site-6:three_site+2]
elif coord_df['strand'] == '-':
s5 = fa_dict[coord_df['chromosome']][five_site-6:five_site+2]
s5 = SP.reverse_complement(s5)
s3 = fa_dict[coord_df['chromosome']][three_site-2:three_site+6]
s3 = SP.reverse_complement(s3)
column_dict['seq5'].append(s5)
column_dict['seq3'].append(s3)
scores = SP.simple_score_junction(s5, s3, pssm)
column_dict['3p score'].append(scores[1])
column_dict['5p score'].append(scores[0])
new_quant_df = quant_df[quant_df.index.isin(new_index)][['genome coord','chromosome',
'strand','transcript','position','type']]
for column, data in column_dict.iteritems():
new_quant_df[column] = data
new_quant_df = new_quant_df.drop_duplicates(subset='genome coord', keep='first').set_index('genome coord')
new_quant_df = SP.backfill_splice_sites(new_quant_df, gff3, fa_dict, pssm, organism=organism)
#for n in range(len(new_quant_df['seq5'].iloc[0])):
# new_quant_df['Base 5-'+str(n)] = [x[n] for x in new_quant_df['seq5']]
#for n in range(len(new_quant_df['seq3'].iloc[0])):
# new_quant_df['Base 3-'+str(n)] = [x[n] for x in new_quant_df['seq3']]
#new_quant_df = new_quant_df.drop(['seq5','seq3'], axis=1)
new_quant_df = SP.find_score_branches_ppy(new_quant_df, '/home/jordan/GENOMES/S288C/S288C_branches2.txt', fa_dict)
return new_quant_df
def build_junction_df(junction_bed, gff3_file, fasta, organism=None):
transcript_dict = SP.build_transcript_dict(gff3_file, organism=organism)
if type(fasta) == str:
fasta=SP.make_fasta_dict(fasta)
junction_dict = build_junction_dict(junction_bed, gff3_file, transcript_dict, organism=organism)
junction_count = 0
for tx, junctions in junction_dict.iteritems():
junction_count += len(junctions)
junction_df = pd.DataFrame(index=range(junction_count), columns=['intron tuple','chromosome','start','end','strand','depth','type','size','annotated intron size','annotated intron start','annotated intron end'])
n=0
for tx, junctions in junction_dict.iteritems():
for junction in junctions:
junction_df.ix[n] = [tx]+junction
n+=1
sequence1 = []
sequence2 = []
ann_seq1 = []
ann_seq2 = []
seq_type1 = []
seq_type2 = []
df_tx = []
for index, row in junction_df.iterrows():
df_tx.append(row['intron tuple'][0])
chrom = convert_chrom(row['chromosome'])
if row['strand'] == '+':
curr1 = fasta[chrom][(row['start']-1):(row['start']+7)]
sequence1.append(curr1)
curr2 = fasta[chrom][(row['end']-5):(row['end']+3)]
sequence2.append(curr2)
if row['annotated intron start'] is None:
ann_seq1.append(None)
ann_seq2.append(None)
else:
ann_seq1.append(fasta[chrom][(row['annotated intron start']-1):(row['annotated intron start']+7)])
ann_seq2.append(fasta[chrom][(row['annotated intron end']-5):(row['annotated intron end']+3)])
elif row['strand'] == '-':
curr1 = SP.reverse_complement(fasta[chrom][(row['start']-6):(row['start']+2)])
sequence1.append(curr1)
curr2 = SP.reverse_complement(fasta[chrom][(row['end']-2):(row['end']+6)])
sequence2.append(curr2)
if row['annotated intron start'] is None:
ann_seq1.append(None)
ann_seq2.append(None)
else:
ann_seq1.append(SP.reverse_complement(fasta[chrom][row['annotated intron start']-6:row['annotated intron start']+2]))
ann_seq2.append(SP.reverse_complement(fasta[chrom][row['annotated intron end']-2:row['annotated intron end']+6]))
else:
sequence1.append('NNNNNNNN')
sequence2.append('NNNNNNNN')
ann_seq1.append('NNNNNNNN')
ann_seq2.append('NNNNNNNN')
if row['type'] == 'Annotated':
seq_type1.append('5p annotated')
seq_type2.append('3p annotated')
elif row['type'] == '5p tethered':
seq_type1.append('5p annotated')
seq_type2.append(curr2[4:6])
else:
seq_type1.append(curr1[2:4])
seq_type2.append(curr2[4:6])
junc_seq_df = junction_df
junc_seq_df['sequence1'] = sequence1
junc_seq_df['sequence2'] = sequence2
junc_seq_df['seq type1'] = seq_type1
junc_seq_df['seq type2'] = seq_type2
junc_seq_df['annotated sequence1'] = ann_seq1
junc_seq_df['annotated sequence2'] = ann_seq2
junc_seq_df['transcript'] = df_tx
return junc_seq_df
