def choose_seqs_to_remove(
chain_ids,
max_hdist=4,
tdbg=False): # choose one of <chain_ids> to eliminate
# look for pairs with the same locus that
ids_to_remove = set(u for u in chain_ids if getloc(u) == '?')
if tdbg and len(
ids_to_remove
) > 0: # i think this actually can't happen a.t.m. TODO maybe remove it
print ' removed %d with missing annotations' % len(
ids_to_remove)
dbgstr = []
n_equivalent = 0
for tpair in itertools.combinations(chain_ids, 2):
if len(set(getloc(u) for u in tpair)) > 1:
continue
if len(set(len(gval(u, 'seqs')) for u in tpair)) > 1:
continue
hdist = utils.hamming_distance(*[gval(u, 'seqs') for u in tpair])
if tdbg:
dbgstr.append(
utils.color('blue' if hdist == 0 else 'yellow',
'%d' % hdist))
if hdist <= max_hdist: # TODO would be nice to be able to combine their sequences, but I think propagating the resulting annotation modifications would be hard
# print ' identical sequence overlap, choosing longer one'
better_id, worse_id = sorted(
tpair, key=lambda q: utils.ambig_frac(gval(q, 'seqs'))
) # TODO if we're tossing one with hdist > 0, maybe should take the lower-shm one if they're the same length?
ids_to_remove.add(worse_id)
n_equivalent += 1
if tdbg and len(dbgstr) > 0:
print ' %d pair%s equivalent with hdists %s' % (
n_equivalent, utils.plural(n_equivalent), ' '.join(dbgstr))
# remove unproductive
dbgstr = []
unproductive_ids = []
for uid in chain_ids:
if not utils.is_functional(
all_antns[uid], all_antns[uid]['unique_ids'].index(uid)):
unproductive_ids.append(uid)
if tdbg:
dbgstr.append(
utils.is_functional_dbg_str(
all_antns[uid],
all_antns[uid]['unique_ids'].index(uid),
sep='+'))
# unproductive_ids = [u for u in chain_ids if not utils.is_functional(all_antns[u], all_antns[u]['unique_ids'].index(u))] # this way is only one line, which may or may not be nicer
if tdbg and len(unproductive_ids) > 0:
print ' %d unproductive %s' % (len(unproductive_ids),
', '.join(dbgstr))
ids_to_remove |= set(unproductive_ids)
return ids_to_remove
def print_seq_in_reco_event(original_line,
iseq,
extra_str='',
label='',
one_line=False,
seed_uid=None,
check_line_integrity=False):
"""
Print ascii summary of recombination event and mutation.
If <one_line>, then skip the germline lines, and only print the final_seq line.
"""
line = original_line
if check_line_integrity: # it's very important not to modify <line> -- this lets you verify that you aren't
line = copy.deepcopy(
original_line) # copy that we can modify without changing <line>
delstrs = {
d: '.' * line[d + '_del']
for d in utils.all_erosions
} # NOTE len(delstrs[<del>]) is not in general the same as len(line[<del>_del])
if len(
delstrs['v_5p']
) > 50: # don't print a million dots if left-side v deletion is really big
delstrs['v_5p'] = '.%d.' % len(delstrs['v_5p'])
# if there isn't enough space for dots in the vj line, we add some dashes to everybody so things fit (rare in heavy chain rearrangements, but pretty common in light chain)
d_plus_inserts_length = len(line['vd_insertion'] + line['d_gl_seq'] +
line['dj_insertion'])
if line['v_3p_del'] + line[
'j_5p_del'] > d_plus_inserts_length: # if dots for v and j interior deletions will be longer than <d_plus_inserts_length>
delstrs['v_3p'] = '.%d.' % line['v_3p_del']
delstrs['j_5p'] = '.%d.' % line['j_5p_del']
gapstr = '-' * (len(delstrs['v_3p'] + delstrs['j_5p']) -
d_plus_inserts_length)
gap_insert_point = len(
line['fv_insertion'] + delstrs['v_5p'] + line['v_gl_seq']
) # it doesn't really matter exactly where we put the blue dashes, as long as it's the same place in all four lines, but this is a good spot
extra_space_because_of_fixed_nospace = max(
0, d_plus_inserts_length - len(delstrs['v_3p'] + delstrs['j_5p'])
) # if shortening the <delstrs> already over-compensated for the lack of space (i.e., if the number of dashes necessary is zero), then we need to add some dots to the vj line below
else:
gapstr = ''
gap_insert_point = None
extra_space_because_of_fixed_nospace = 0
eroded_seqs_dots = {
r: delstrs[r + '_5p'] + line[r + '_gl_seq'] + delstrs[r + '_3p']
for r in utils.regions
}
# build the three germline lines
insert_line = ' ' * (len(line['fv_insertion']) + line['lengths']['v'] + len(delstrs['v_5p'])) \
+ line['vd_insertion'] + ' ' * line['lengths']['d'] + line['dj_insertion'] \
+ ' ' * (line['lengths']['j'] + line['j_3p_del'] + len(line['jf_insertion']))
germline_d_start = len(line['fv_insertion']) + line['lengths']['v'] + len(
line['vd_insertion']) - line['d_5p_del']
germline_d_end = germline_d_start + line['d_5p_del'] + line['lengths'][
'd'] + line['d_3p_del']
d_line = ' ' * (germline_d_start + len(delstrs['v_5p'])) \
+ eroded_seqs_dots['d'] \
+ ' ' * (len(line['j_gl_seq']) + len(line['dj_insertion']) - line['d_3p_del'] + line['j_3p_del'] + len(line['jf_insertion']))
germline_v_end = len(line['fv_insertion']) + len(line['v_gl_seq']) + line[
'v_3p_del'] - 1 # position in the query sequence at which we find the last base of the v match. NOTE we subtract off the v_5p_del because we're *not* adding dots for that deletion (it's just too long)
germline_j_start = germline_d_end + 1 - line['d_3p_del'] + len(
line['dj_insertion']) - line['j_5p_del']
vj_line = ' ' * len(line['fv_insertion']) + eroded_seqs_dots['v'] + '.' * extra_space_because_of_fixed_nospace \
+ ' ' * (germline_j_start - germline_v_end - 2) + eroded_seqs_dots['j'] + ' ' * len(line['jf_insertion'])
# and the query line
qrseq_line = ' ' * len(
delstrs['v_5p']) + line['seqs'][iseq] + ' ' * line['j_3p_del']
outstrs = [insert_line, d_line, vj_line, qrseq_line]
check_outsr_lengths(
line, outstrs, fix=True
) # I think the only way they can be different is if the d right side erosion is so long that it hangs over the right side of the j
if gap_insert_point is not None:
for istr in [
0, 1, 3
]: # everybody except the vj line, which already has the modified interior delstrs above
outstrs[
istr] = outstrs[istr][:gap_insert_point] + gapstr + outstrs[
istr][gap_insert_point:]
check_outsr_lengths(line, outstrs, fix=True)
colors = [[[] for _ in range(len(ostr))] for ostr in outstrs]
if indelutils.has_indels(line['indelfos'][iseq]):
# outstrs, colors = old_indel_shenanigans(line, iseq, outstrs, colors)
outstrs, colors = indel_shenanigans(line, iseq, outstrs, colors)
outstrs = add_colors(outstrs, colors, line)
suffixes = [
'insert%s\n' %
('s' if utils.has_d_gene(utils.get_locus(line['v_gene'])) else ''),
'%s\n' % (utils.color_gene(line['d_gene'])),
'%s %s\n' %
(utils.color_gene(line['v_gene']), utils.color_gene(line['j_gene'])),
'%s %4.2f mut %s\n' %
(get_uid_str(line, iseq, seed_uid), line['mut_freqs'][iseq],
utils.color('red', utils.is_functional_dbg_str(line, iseq)))
]
outstrs = [
'%s%s %s' % (extra_str, ostr, suf)
for ostr, suf in zip(outstrs, suffixes)
]
if label != '': # this doesn't really work if the edge of the removed string is the middle of a color code... but oh well, it doesn't really happen any more since I shortened the kbound label from waterer.py
offset = max(
0,
len(extra_str) -
2) # skootch <label> this many positions leftward into <extra_str>
removed_str = outstrs[0][offset:offset +
utils.len_excluding_colors(label)]
outstrs[0] = outstrs[0][:offset] + label + outstrs[0][
utils.len_excluding_colors(label) +
offset:] # NOTE this *replaces* the bases in <extra_str> with <label>, which is only fine if they're spaces
if removed_str.strip() != '':
print '%s%s (covered by label \'%s\')' % (
' ' * offset, utils.color('red', removed_str), label)
if one_line:
outstrs = outstrs[-1:] # remove all except the query seq line
elif not utils.has_d_gene(utils.get_locus(line['v_gene'])):
outstrs.pop(1) # remove the d germline line
print ''.join(outstrs),
if check_line_integrity:
if set(line.keys()) != set(original_line.keys()):
raise Exception('ack 1')
for k in line:
if line[k] != original_line[k]:
print 'key %s differs:\n %s\n %s ' % (k, line[k],
original_line[k])
raise Exception('')