def plot_map(fasta_filename,
c_filename,
factor,
c2_filename='',
psipred_filename='',
pdb_filename='',
is_heavy=False,
chain='',
sep=' '):
acc = c_filename.split('.')[0]
### get sequence
seq = parse_fasta.read_fasta(open(fasta_filename, 'r')).values()[0][0]
ref_len = len(seq)
### get top "factor" * "ref_len" predicted contacts
contacts = parse_contacts.parse(open(c_filename, 'r'), sep)
contacts_x = []
contacts_y = []
scores = []
contact_dict = {}
count = 0
for i in range(len(contacts)):
score = contacts[i][0]
c_x = contacts[i][1] - 1
c_y = contacts[i][2] - 1
pos_diff = abs(c_x - c_y)
too_close = pos_diff < 5
if not too_close:
contacts_x.append(c_x)
contacts_y.append(c_y)
scores.append(score)
count += 1
if count >= ref_len * factor:
break
### start plotting
fig = plt.figure(figsize=(8, 8), dpi=100)
ax = fig.add_subplot(111)
#ax = plt.axes([.1, .1, .8, .8], frameon=False)
ax.set_xlim(xmin=-1)
ax.set_ylim(ymin=-1)
### plot secondary structure on the diagonal if given
if psipred_filename:
ss = parse_psipred.horizontal(open(psipred_filename, 'r'))
for i in range(len(ss)):
if ss[i] == 'H':
plt.plot(i, i, 'o', c='#8B0043',
mec="#8B0043") #, markersize=8)
if ss[i] == 'E':
plt.plot(i, i, 'D', c='#0080AD',
mec="#0080AD") #, markersize=8)
if ss[i] == 'C':
plt.plot(i, i, 'D', c='#CCCCCC', mec="#CCCCCC", markersize=4)
### plot reference contacts in the background if given
if pdb_filename:
res_lst = parse_pdb.get_coordinates(open(pdb_filename, 'r'), chain)
cb_lst = parse_pdb.get_cb_coordinates(open(pdb_filename, 'r'), chain)
atom_seq = parse_pdb.get_atom_seq(open(pdb_filename, 'r'), chain)
align = pairwise2.align.globalms(atom_seq, seq, 2, -1, -0.5, -0.1)
atom_seq_ali = align[-1][0]
seq_ali = align[-1][1]
j = 0
gapped_res_lst = []
gapped_cb_lst = []
for i in xrange(len(atom_seq_ali)):
if atom_seq_ali[i] == '-':
gapped_res_lst.append('-')
gapped_cb_lst.append('-')
elif seq_ali[i] == '-':
j += 1
continue
else:
gapped_res_lst.append(res_lst[j])
gapped_cb_lst.append(cb_lst[j])
j += 1
if is_heavy:
dist_mat = get_heavy_contacts(gapped_res_lst)
heavy_cutoff = 5
ref_contact_map = dist_mat < heavy_cutoff
ref_contacts = np.where(dist_mat < heavy_cutoff)
else:
dist_mat = get_cb_contacts(gapped_cb_lst)
cb_cutoff = 8
ref_contact_map = dist_mat < cb_cutoff
ref_contacts = np.where(dist_mat < cb_cutoff)
ref_contacts_x = ref_contacts[0]
ref_contacts_y = ref_contacts[1]
PPVs = get_ppvs(contacts_x, contacts_y, ref_contact_map, atom_seq_ali,
ref_len, factor)
tp_colors = get_tp_colors(contacts_x, contacts_y, ref_contact_map,
atom_seq_ali)
print '%s\t%s' % (acc, PPVs[-1])
ax.scatter(ref_contacts_x,
ref_contacts_y,
marker='o',
c='#DDDDDD',
lw=0)
### plot predicted contacts from second contact map if given
if c2_filename:
contacts2 = parse_contacts.parse(open(c2_filename, 'r'), sep)
contacts2_x = []
contacts2_y = []
scores2 = []
contact_dict2 = {}
count = 0
for i in range(len(contacts2)):
score = contacts2[i][0]
c_x = contacts2[i][1] - 1
c_y = contacts2[i][2] - 1
pos_diff = abs(c_x - c_y)
too_close = pos_diff < 5
if not too_close:
contacts2_x.append(c_x)
contacts2_y.append(c_y)
scores2.append(score)
count += 1
if count >= ref_len * factor:
break
### use TP/FP color coding if reference contacts given
if pdb_filename:
PPVs2 = get_ppvs(contacts2_x, contacts2_y, ref_contact_map,
atom_seq_ali, ref_len, factor)
tp2_colors = get_tp_colors(contacts2_x, contacts2_y,
ref_contact_map, atom_seq_ali)
print '%s\t%s' % (acc, PPVs2[-1])
fig.suptitle(
'%s\n%s (upper left) PPV = %.2f | %s (lower right) PPV = %.2f'
% (acc, c_filename, PPVs[-1], c2_filename, PPVs2[-1]))
sc = ax.scatter(contacts2_y[::-1],
contacts2_x[::-1],
marker='o',
c=tp2_colors[::-1],
linewidths=0.0)
sc = ax.scatter(contacts_x[::-1],
contacts_y[::-1],
marker='o',
c=tp_colors[::-1],
linewidths=0.0)
else:
fig.suptitle('%s\n%s (upper left) | %s (lower right)' %
(acc, c_filename, c2_filename))
sc = ax.scatter(contacts2_y[::-1],
contacts2_x[::-1],
marker='o',
c='#D70909',
edgecolor='#D70909',
s=8,
linewidths=0.5)
sc = ax.scatter(contacts_x[::-1],
contacts_y[::-1],
marker='o',
c='#004F9D',
edgecolor='#004F9D',
s=8,
linewidths=0.5)
### plot predicted contacts from first contact map on both triangles
### if no second contact map given
else:
if pdb_filename:
fig.suptitle('%s (%s)\nPPV = %.2f' % (acc, c_filename, PPVs[-1]))
sc = ax.scatter(contacts_x[::-1],
contacts_y[::-1],
marker='o',
c=tp_colors[::-1],
linewidths=0.0)
sc = ax.scatter(contacts_y[::-1],
contacts_x[::-1],
marker='o',
c=tp_colors[::-1],
linewidths=0.0)
else:
#fig.suptitle('%s (%s)' % (acc, c_filename))
#sc = ax.scatter(contacts_x[::-1], contacts_y[::-1], marker='o', c=scores[::-1], s=6, alpha=0.75, cmap=cm.jet, linewidths=0.5)
#sc = ax.scatter(contacts_y[::-1], contacts_x[::-1], marker='o', c=scores[::-1], s=6, alpha=0.75, cmap=cm.jet, linewidths=0.5)
sc = ax.scatter(contacts_x[::-1],
contacts_y[::-1],
marker='o',
c='#004F9D',
edgecolor='#004F9D',
s=8,
linewidths=0.5)
sc = ax.scatter(contacts_y[::-1],
contacts_x[::-1],
marker='o',
c='#004F9D',
edgecolor='#004F9D',
s=8,
linewidths=0.5)
#plt.colorbar(sc)
plt.gca().set_xlim([-1, ref_len])
plt.gca().set_ylim([-1, ref_len])
plt.savefig('%s.cm.png' % c_filename, bbox_inches=0)
def plot_map(fasta_filename, c_filename, factor, c2_filename='', psipred_filename='', pdb_filename='', is_heavy=False, chain='', sep=' '):
acc = c_filename.split('.')[0]
### get sequence
seq = parse_fasta.read_fasta(open(fasta_filename, 'r')).values()[0][0]
ref_len = len(seq)
### get top "factor" * "ref_len" predicted contacts
contacts = parse_contacts.parse(open(c_filename, 'r'), sep)
contacts_x = []
contacts_y = []
scores = []
contact_dict = {}
count = 0
for i in range(len(contacts)):
score = contacts[i][0]
c_x = contacts[i][1] - 1
c_y = contacts[i][2] - 1
pos_diff = abs(c_x - c_y)
too_close = pos_diff < 5
if not too_close:
contacts_x.append(c_x)
contacts_y.append(c_y)
scores.append(score)
count += 1
if count >= ref_len * factor:
break
### start plotting
fig = plt.figure(figsize=(8,8), dpi=100)
ax = fig.add_subplot(111)
#ax = plt.axes([.1, .1, .8, .8], frameon=False)
ax.set_xlim(xmin=-1)
ax.set_ylim(ymin=-1)
### plot secondary structure on the diagonal if given
if psipred_filename:
ss = parse_psipred.horizontal(open(psipred_filename, 'r'))
for i in range(len(ss)):
if ss[i] == 'H':
plt.plot(i, i, 'o', c='#8B0043', mec="#8B0043")#, markersize=8)
if ss[i] == 'E':
plt.plot(i, i, 'D', c='#0080AD', mec="#0080AD")#, markersize=8)
if ss[i] == 'C':
plt.plot(i, i, 'D', c='#CCCCCC', mec="#CCCCCC", markersize=4)
### plot reference contacts in the background if given
if pdb_filename:
res_lst = parse_pdb.get_coordinates(open(pdb_filename, 'r'), chain)
cb_lst = parse_pdb.get_cb_coordinates(open(pdb_filename, 'r'), chain)
atom_seq = parse_pdb.get_atom_seq(open(pdb_filename, 'r'), chain)
align = pairwise2.align.globalms(atom_seq, seq, 2, -1, -0.5, -0.1)
atom_seq_ali = align[-1][0]
seq_ali = align[-1][1]
j = 0
gapped_res_lst = []
gapped_cb_lst = []
for i in xrange(len(atom_seq_ali)):
if atom_seq_ali[i] == '-':
gapped_res_lst.append('-')
gapped_cb_lst.append('-')
elif seq_ali[i] == '-':
j += 1
continue
else:
gapped_res_lst.append(res_lst[j])
gapped_cb_lst.append(cb_lst[j])
j += 1
if is_heavy:
dist_mat = get_heavy_contacts(gapped_res_lst)
heavy_cutoff = 5
ref_contact_map = dist_mat < heavy_cutoff
ref_contacts = np.where(dist_mat < heavy_cutoff)
else:
dist_mat = get_cb_contacts(gapped_cb_lst)
cb_cutoff = 8
ref_contact_map = dist_mat < cb_cutoff
ref_contacts = np.where(dist_mat < cb_cutoff)
ref_contacts_x = ref_contacts[0]
ref_contacts_y = ref_contacts[1]
PPVs = get_ppvs(contacts_x, contacts_y, ref_contact_map, atom_seq_ali, ref_len, factor)
tp_colors = get_tp_colors(contacts_x, contacts_y, ref_contact_map, atom_seq_ali)
print '%s\t%s' % (acc, PPVs[-1])
ax.scatter(ref_contacts_x, ref_contacts_y, marker='o', c='#DDDDDD', lw=0)
### plot predicted contacts from second contact map if given
if c2_filename:
contacts2 = parse_contacts.parse(open(c2_filename, 'r'), sep)
contacts2_x = []
contacts2_y = []
scores2 = []
contact_dict2 = {}
count = 0
for i in range(len(contacts2)):
score = contacts2[i][0]
c_x = contacts2[i][1] - 1
c_y = contacts2[i][2] - 1
pos_diff = abs(c_x - c_y)
too_close = pos_diff < 5
if not too_close:
contacts2_x.append(c_x)
contacts2_y.append(c_y)
scores2.append(score)
count += 1
if count >= ref_len * factor:
break
### use TP/FP color coding if reference contacts given
if pdb_filename:
PPVs2 = get_ppvs(contacts2_x, contacts2_y, ref_contact_map, atom_seq_ali, ref_len, factor)
tp2_colors = get_tp_colors(contacts2_x, contacts2_y, ref_contact_map, atom_seq_ali)
print '%s\t%s' % (acc, PPVs2[-1])
fig.suptitle('%s\n%s (upper left) PPV = %.2f | %s (lower right) PPV = %.2f' % (acc, c_filename, PPVs[-1], c2_filename, PPVs2[-1]))
sc = ax.scatter(contacts2_y[::-1], contacts2_x[::-1], marker='o', c=tp2_colors[::-1], linewidths=0.0)
sc = ax.scatter(contacts_x[::-1], contacts_y[::-1], marker='o', c=tp_colors[::-1], linewidths=0.0)
else:
fig.suptitle('%s\n%s (upper left) | %s (lower right)' % (acc, c_filename, c2_filename))
sc = ax.scatter(contacts2_y[::-1], contacts2_x[::-1], marker='o', c='#D70909', edgecolor='#D70909', s=8, linewidths=0.5)
sc = ax.scatter(contacts_x[::-1], contacts_y[::-1], marker='o', c='#004F9D', edgecolor='#004F9D', s=8, linewidths=0.5)
### plot predicted contacts from first contact map on both triangles
### if no second contact map given
else:
if pdb_filename:
fig.suptitle('%s (%s)\nPPV = %.2f' % (acc, c_filename, PPVs[-1]))
sc = ax.scatter(contacts_x[::-1], contacts_y[::-1], marker='o', c=tp_colors[::-1], linewidths=0.0)
sc = ax.scatter(contacts_y[::-1], contacts_x[::-1], marker='o', c=tp_colors[::-1], linewidths=0.0)
else:
fig.suptitle('%s (%s)' % (acc, c_filename))
#sc = ax.scatter(contacts_x[::-1], contacts_y[::-1], marker='o', c=scores[::-1], s=6, alpha=0.75, cmap=cm.jet, linewidths=0.5)
#sc = ax.scatter(contacts_y[::-1], contacts_x[::-1], marker='o', c=scores[::-1], s=6, alpha=0.75, cmap=cm.jet, linewidths=0.5)
sc = ax.scatter(contacts_x[::-1], contacts_y[::-1], marker='o', c='#004F9D', edgecolor='#004F9D', s=8, linewidths=0.5)
sc = ax.scatter(contacts_y[::-1], contacts_x[::-1], marker='o', c='#004F9D', edgecolor='#004F9D', s=8, linewidths=0.5)
#plt.colorbar(sc)
plt.gca().set_xlim([-1,ref_len])
plt.gca().set_ylim([-1,ref_len])
plt.savefig('%s.cm.png' % c_filename, bbox_inches=0)
def plot_map(fasta_filename,
c_filename,
factor=1.0,
th=-1,
c2_filename='',
psipred_horiz_fname='',
psipred_vert_fname='',
iupred_fname='',
pdb_filename='',
is_heavy=False,
chain='',
sep=',',
outfilename='',
ali_filename='',
meff_filename='',
name='',
start=0,
end=-1):
#acc = c_filename.split('.')[0]
#acc = fasta_filename.split('.')[0][:4]
if name == '':
acc = '.'.join(os.path.basename(fasta_filename).split('.')[:-1])
else:
acc = name
### get sequence
seq = parse_fasta.read_fasta(open(fasta_filename, 'r')).values()[0][0]
ref_len = len(seq)
### trim sequence according to given positions
### default: take full sequence
if end == -1:
end = ref_len
seq = seq[start:end]
ref_len = len(seq)
unit = (ref_len / 50.0)
if ali_filename:
coverage_lst = get_ali_coverage(ali_filename)
max_cover = max(coverage_lst)
# Current Meff format does not work
# elif meff_filename:
# coverage_lst = get_meff_coverage(meff_filename)
# max_cover = max(coverage_lst)
else:
coverage_lst = np.zeros(ref_len)
max_cover = 0
average_disorder = 0.
average_order = 0.
fraction_disorder = 0.
cover_order = 0.
cover_disorder = 0.
if iupred_fname:
disorder = parse_iupred.pred(open(iupred_fname, 'r'))
else:
disorder = np.zeros(ref_len)
average_disorder = np.sum(disorder) / ref_len
fraction_disorder = 0.0
num_disorder = 0
num_order = 0
j = 0
for i in disorder:
if (i > 0.5):
fraction_disorder += 1 / ref_len
num_disorder += 1
cover_disorder += coverage_lst[j]
else:
num_order += 1
cover_order += coverage_lst[j]
j += 1
if (num_disorder > 0):
cover_disorder = cover_disorder / num_disorder
if (num_order > 0):
cover_order = cover_order / num_order
### get top "factor" * "ref_len" predicted contacts
contacts = parse_contacts.parse(open(c_filename, 'r'), sep, 1)
contacts_np = parse_contacts.get_numpy_cmap(contacts)
contacts_np = contacts_np[start:end, start:end]
contacts_x = []
contacts_y = []
scores = []
mixscores = []
disoscores = []
orderscores = []
tooclose = []
contact_dict = {}
if iupred_fname:
disorder = parse_iupred.pred(open(iupred_fname, 'r'))
else:
disorder = np.zeros(ref_len)
count = 1.e-20
mixcount = 1.e-20
disocount = 1.e-20
ordercount = 1.e-20
longcount = 1.e-20
highscore = 0
numbins = 20
sum = 0.0
longsum = 0.0
disosum = 0.0
ordersum = 0.0
mixsum = 0.0
average = 0.0
longaverage = 0.0
mixaverage = 0.0
disoaverage = 0.0
orderaverage = 0.0
histo = np.zeros(numbins)
disotop = 0
ordertop = 0
doubletop = 0
mixcount = 0
mixtop = 0
separation = 0.0
# We actually divide the analysis into three groups (ordered,disordered and mixed)
for i in range(len(contacts)):
score = contacts[i][0]
c_x = contacts[i][1] - 1
c_y = contacts[i][2] - 1
# only look at contacts within given range
# default: take full sequence range into account
if c_x < start or c_x >= end:
continue
if c_y < start or c_y >= end:
continue
pos_diff = abs(c_x - c_y)
too_close = pos_diff < 5
long_dist = pos_diff > 24
if not too_close:
if score > th:
contacts_x.append(c_x - start)
contacts_y.append(c_y - start)
if (disorder[c_x] > 0.5 and disorder[c_y] > 0.5):
disocount += 1
disoscores.append(score)
if (disocount <= ref_len * factor):
disosum += score
disoaverage = disosum / disocount
elif (disorder[c_x] > 0.5 or disorder[c_y] > 0.5):
mixcount += 1
mixscores.append(score)
if (mixcount <= ref_len * factor):
mixsum += score
mixaverage = mixsum / mixcount
else:
ordercount += 1
orderscores.append(score)
if (ordercount <= ref_len * factor):
ordersum += score
orderaverage = ordersum / ordercount
count += 1
scores.append(score)
if (count <= ref_len * factor):
sum += score
average = sum / count
separation += pos_diff
if long_dist:
longcount += 1
if (longcount <= ref_len * factor):
longsum += score
longaverage = longsum / longcount
else:
tooclose.append(score)
# statline="Highs: %.1f (%.1f%%) (%.1f%%) average: %.2f (%.2f) (%.2f) Meff: %.0f Diso: %.1f%% " % (count/ref_len,100*mixcount/count,100*disocount/count,average,mixaverage,disoaverage,max_cover,100*fraction_disorder)
# statline="Highs: %.1f %.3f %.3f average: %.2f %.2f %.2f Meff: %.0f Diso: %.3f " % (count/ref_len,mixcount/count,disocount/count,average,mixaverage,disoaverage,max_cover,fraction_disorder)
# statline="Length: %d NumAli: %d Counts: %d %d %d %.3f %.3f %.3f %.3f\n" % ( ref_len,max_cover,(count-mixcount-disocount),mixcount,disocount,sum,mixsum,disosum,fraction_disorder)
statline = "NumAli: %d %d %d Length: %d %d %d Counts: %d %d %d %d RelContacts: %.3f %.3f %.3f Disorder: %.3f Long: %.3f %.3f %.3f \n" % (
max_cover, cover_order, cover_disorder, ref_len, num_order,
num_disorder, count, ordercount, mixcount, disocount, count /
(ref_len + 1.e-20), ordercount / (1.e-20 + num_order), disocount /
(1.e-20 + num_disorder), fraction_disorder, longcount / count,
longcount / (ref_len + 1.e-20), separation / count)
statfig = plt.figure(figsize=(8, 8), dpi=96, facecolor='w')
plt.hist((tooclose, scores),
numbins,
range=(0, 1),
histtype='bar',
normed=(numbins, numbins),
alpha=0.75,
label=['Too_Close', 'Contacts'])
plt.xlabel('Score')
plt.ylabel('Normalized count')
statfig.suptitle('%s\n%s\n' % (c_filename, line))
### start plotting
fig = plt.figure(figsize=(8, 8), dpi=96, facecolor='w')
ax = fig.add_subplot(111) #, aspect='auto')
ax.set_adjustable('box-forced')
ax.tick_params(direction='out', right='off', top='off')
ax.set_xlim([-unit, ref_len])
ax.set_ylim([-unit, ref_len])
max_cover = 0
### plot alignment coverage if alignemnt given (only on Y-axis)
if ali_filename: # or meff_filename:
# adjust overall canvas
ax = plt.subplot2grid((8, 8), (1, 1), colspan=7,
rowspan=7) #, aspect='auto')
#ax.set_adjustable('box-forced')
#ax.set_autoscale_on(False)
ax.autoscale(False)
ax.tick_params(direction='out',
labelleft='off',
right='off',
top='off')
ax.set_xlim([-unit, ref_len])
ax.set_ylim([-unit, ref_len])
if ali_filename:
coverage_lst = get_ali_coverage(ali_filename)
#elif meff_filename:
# coverage_lst = get_meff_coverage(meff_filename)
max_cover = max(coverage_lst)
#lt = pow(10, max(1,floor(log10(max_cover)) - 1))
#upper = int(ceil(max_cover/float(lt)) * lt)
ax2 = plt.subplot2grid((8, 8), (1, 0), rowspan=7, sharey=ax)
#ax2.set_adjustable('box-forced')
#ax2.set_autoscale_on(False)
ax2.autoscale(False)
#print len([0]+coverage_lst+[0])
#print len([0]+range(ref_len)+[ref_len-1])
ax2.plot([0] + coverage_lst + [0],
[0] + range(ref_len) + [ref_len - 1],
'k',
lw=0)
ax2.axvline(x=max_cover * 0.25, lw=0.5, c='black', ls=':')
ax2.axvline(x=max_cover * 0.5, lw=0.5, c='black', ls=':')
ax2.axvline(x=max_cover * 0.75, lw=0.5, c='black', ls=':')
ax2.fill([0] + coverage_lst + [0],
[0] + range(ref_len) + [ref_len - 1],
facecolor='gray',
lw=0,
alpha=0.5)
ax2.set_xticks([0, max_cover])
ax2.tick_params(axis='x', top='off', direction='out')
ax2.invert_xaxis()
#ax2.spines['top'].set_visible(False)
#ax2.spines['left'].set_visible(False)
#ax.get_xaxis().tick_bottom()
#ax.get_yaxis().tick_right()
ax2.grid()
ax2.set_ylim([-unit, ref_len])
#ax3 = plt.subplot2grid((8,8), (0,1), colspan=7, sharex=ax)
#ax3.set_adjustable('box-forced')
#ax3.set_autoscale_on(False)
#ax3.autoscale(False)
#ax3.plot([0]+range(ref_len)+[ref_len-1], [0]+coverage_lst+[0], 'k', lw=0)
#ax3.axhline(y=max_cover*0.25, lw=0.5, c='black', ls=':')
#ax3.axhline(y=max_cover*0.5, lw=0.5, c='black', ls=':')
#ax3.axhline(y=max_cover*0.75, lw=0.5, c='black', ls=':')
#ax3.fill([0]+range(ref_len)+[ref_len-1], [0]+coverage_lst+[0], facecolor='gray', lw=0, alpha=0.5)
#ax3.xaxis.tick_top()
#ax3.set_yticks([0, max_cover])
#ax3.tick_params(labelbottom='off')
ax2.tick_params(axis='y', right='off', direction='out', left='on')
#ax3.spines['top'].set_visible(False)
#ax3.spines['right'].set_visible(False)
#ax.get_xaxis().tick_top()
#ax.get_yaxis().tick_left()
#ax3.grid()
#ax3.set_xlim([-unit,ref_len])
### plot secondary structure along axis if given
average_disorder = 0.
fraction_disorder = 0.
#statline = "Highs: %.1f Aver: %.2f Meff: %.0f" % (count/ref_len,average,max_cover)
ax.get_xaxis().tick_top()
ax.get_yaxis().tick_left()
if iupred_fname:
ax = plt.subplot2grid((8, 8), (1, 1), colspan=7,
rowspan=7) #, aspect='auto')
#ax.set_adjustable('box-forced')
#ax.set_autoscale_on(False)
ax.autoscale(False)
ax.tick_params(direction='out',
labelleft='off',
right='off',
top='off')
ax.set_xlim([-unit, ref_len])
ax.set_ylim([-unit, ref_len])
average_disorder = np.sum(disorder) / ref_len
fraction_disorder = 0.0
for d in disorder:
if (d > 0.5):
fraction_disorder += 1 / ref_len
ax3 = plt.subplot2grid((8, 8), (0, 1), colspan=7, sharex=ax)
ax3.set_adjustable('box-forced')
ax3.set_autoscale_on(False)
ax3.autoscale(False)
ax3.plot([0] + range(ref_len) + [ref_len - 1], [0] + disorder + [0],
'b',
lw=2)
ax3.axhline(y=0.5, lw=0.5, c='black', ls=':')
#ax3.fill([0]+range(ref_len)+[ref_len-1], [0]+disorder+[0], facecolor='gray', lw=0, alpha=0.5)
ax3.xaxis.tick_top()
ax3.set_yticks([0, 1])
ax3.tick_params(labelbottom='off')
ax3.spines['top'].set_visible(False)
ax3.spines['right'].set_visible(False)
ax3.grid()
ax3.set_xlim([-unit, ref_len])
#statline = "Highs: %.3f %.3f %.3f Aver: %.2f Meff: %.0f Diso: %.3f " % (count/ref_len,disocount/count,doublecount/count,average,max_cover,fraction_disorder)
print "STATs: %s %s" % (c_filename, statline)
if psipred_horiz_fname or psipred_vert_fname:
if psipred_horiz_fname:
ss = parse_psipred.horizontal(open(psipred_horiz_fname, 'r'))
else:
ss = parse_psipred.vertical(open(psipred_vert_fname, 'r'))
ss = ss[start:end]
assert len(ss) == ref_len
ax.axhline(y=0, lw=1, c='black')
ax.axvline(x=0, lw=1, c='black')
for i in range(len(ss)):
if ss[i] == 'H':
#ax.plot(-unit/2, i, 's', c='#8B0043', mec="#8B0043")#, markersize=2)
#ax.plot(i, -unit/2, 's', c='#8B0043', mec="#8B0043")#, markersize=2)
#ax.plot(i, -unit/2, 's', c='#8B0043', mec="#8B0043")#, markersize=2)
ax.add_patch(
plt.Rectangle((-unit, i - 0.5),
unit,
1,
edgecolor='#8B0043',
facecolor="#8B0043"))
ax.add_patch(
plt.Rectangle((i - 0.5, -unit),
1,
unit,
edgecolor='#8B0043',
facecolor="#8B0043"))
if ss[i] == 'E':
ax.add_patch(
plt.Rectangle((-unit, i - 0.5),
unit,
1,
edgecolor='#0080AD',
facecolor="#0080AD"))
ax.add_patch(
plt.Rectangle((i - 0.5, -unit),
1,
unit,
edgecolor='#0080AD',
facecolor="#0080AD"))
#ax.plot(-unit/2, i, 's', c='#0080AD', mec="#0080AD")#, markersize=2)
#ax.plot(i, -unit/2, 's', c='#0080AD', mec="#0080AD")#, markersize=2)
if ss[i] == 'C':
continue
### plot reference contacts in the background if given
if pdb_filename:
chain = '*'
# We try to get all chains...
res_lst = parse_pdb.get_coordinates(open(pdb_filename, 'r'), chain)
# cb_lst = parse_pdb.get_ca_coordinates(open(pdb_filename, 'r'), chain)
cb_lst = parse_pdb.get_cb_coordinates(open(pdb_filename, 'r'), chain)
atom_seq = parse_pdb.get_atom_seq(open(pdb_filename, 'r'), chain)
# align = pairwise2.align.globalms(atom_seq, seq, 2, -1, -0.5, -0.1)
align = pairwise2.align.globalms(atom_seq, seq, 2, -1, -10.5, -10.1)
atom_seq_ali = align[-1][0]
seq_ali = align[-1][1]
print atom_seq_ali
print seq_ali
j = 0
gapped_res_lst = []
gapped_cb_lst = []
for i in xrange(len(atom_seq_ali)):
# print i,j
if atom_seq_ali[i] == '-':
gapped_res_lst.append('-')
gapped_cb_lst.append('-')
elif seq_ali[i] == '-':
j += 1
continue
else:
gapped_res_lst.append(res_lst[j])
gapped_cb_lst.append(cb_lst[j])
j += 1
if is_heavy:
dist_mat = get_heavy_contacts(gapped_res_lst)
heavy_cutoff = 5
ref_contact_map = dist_mat < heavy_cutoff
ref_contacts = np.where(dist_mat < heavy_cutoff)
else:
dist_mat = get_cb_contacts(gapped_cb_lst)
cb_cutoff = 8
ref_contact_map = dist_mat < cb_cutoff
ref_contacts = np.where(dist_mat < cb_cutoff)
#ref_contacts = np.where(np.ma.array(dist_mat, mask=np.tri(dist_mat.shape[0]), fill_value=float("inf")) < cb_cutoff)
ref_contacts_x = ref_contacts[0]
ref_contacts_y = ref_contacts[1]
PPVs, TPs, FPs, orderPPVs, orderTPs, orderFPs, mixPPVs, mixTPs, mixFPs, disoPPVs, disoTPs, disoFPs = get_ppvs(
contacts_x, contacts_y, ref_contact_map, atom_seq_ali, ref_len,
factor, disorder)
tp_colors = get_tp_colors(contacts_x, contacts_y, ref_contact_map,
atom_seq_ali)
img = get_colors(contacts_np,
ref_contact_map=dist_mat,
atom_seq_ali=atom_seq_ali,
th=th,
factor=factor)
sc = ax.imshow(img, interpolation='none')
#
print 'PPV: %s %s %s %s %s' % (acc, PPVs[-1], orderPPVs[-1],
mixPPVs[-1], disoPPVs[-1])
cmap = cm.get_cmap("binary")
cmap.set_bad([1, 1, 1, 0])
dist_mat_masked = np.ma.array(dist_mat,
mask=np.tri(dist_mat.shape[0], k=-1))
#sc = ax.imshow(s_score_vec(dist_mat_masked, 5), cmap=cmap, interpolation='none')
ref_contacts_diag_x = []
ref_contacts_diag_y = []
for i in range(len(ref_contacts_x)):
x_i = ref_contacts_x[i]
y_i = ref_contacts_y[i]
if not dist_mat_masked.mask[x_i, y_i] and abs(x_i - y_i) >= 5:
ref_contacts_diag_x.append(x_i)
ref_contacts_diag_y.append(y_i)
#ax.scatter(ref_contacts_diag_x, ref_contacts_diag_y, marker='+', c='#000000')
### plot predicted contacts from second contact map if given
if c2_filename:
contacts2 = parse_contacts.parse(open(c2_filename, 'r'), sep)
contacts2_x = []
contacts2_y = []
scores2 = []
contact_dict2 = {}
count = 0
for i in range(len(contacts2)):
score = contacts2[i][0]
c_x = contacts2[i][1] - 1
c_y = contacts2[i][2] - 1
pos_diff = abs(c_x - c_y)
too_close = pos_diff < 5
if not too_close:
contacts2_x.append(c_x)
contacts2_y.append(c_y)
scores2.append(score)
count += 1
if count >= ref_len * factor:
break
### use TP/FP color coding if reference contacts given
if pdb_filename:
PPVs, TPs, FPs, orderPPVs, orderTPs, orderFPs, mixPPVs, mixTPs, mixFPs, disoPPVs, disoTPs, disoFPs = get_ppvs(
contacts_x, contacts_y, ref_contact_map, atom_seq_ali, ref_len,
factor, disorder)
tp2_colors = get_tp_colors(contacts2_x, contacts2_y,
ref_contact_map, atom_seq_ali)
print '%s %s %s %s' % (acc, PPVs2[-1], TPs2[-1], FPs2[-1])
fig.suptitle(
'%s\nPPV (upper left) = %.2f | PPV (lower right) = %.2f' %
(acc, PPVs[-1], PPVs2[-1]))
sc = ax.scatter(contacts2_y[::-1],
contacts2_x[::-1],
marker='o',
c=tp2_colors[::-1],
s=6,
alpha=0.75,
lw=0)
sc = ax.scatter(contacts_x[::-1],
contacts_y[::-1],
marker='o',
c=tp_colors[::-1],
s=6,
alpha=0.75,
lw=0)
else:
sc = ax.scatter(contacts2_y[::-1],
contacts2_x[::-1],
marker='0',
c='#D70909',
edgecolor='#D70909',
s=6,
linewidths=0.5)
sc = ax.scatter(contacts_x[::-1],
contacts_y[::-1],
marker='o',
c='#004F9D',
edgecolor='#004F9D',
s=6,
linewidths=0.5)
# ppv='PPV: %.2f %.2f %.2f' % (float(PPVs[-1]), float(TPs[-1]), float(FPs[-1]))
ppv = 'PPV: %.2f (%d) %.2f (%d) %.2f (%d) ' % (float(
PPVs[-1]), len(PPVs), float(
mixPPVs[-1]), len(mixPPVs), float(disoPPVs[-1]), len(disoPPVs))
else:
if pdb_filename:
pdb_acc = parse_pdb.get_acc(open(pdb_filename))
if pdb_acc:
if chain:
fig.suptitle(
'%s (PDB: %s, chain %s)\nPPV = %.2f\n%s' %
(c_filename, pdb_acc, chain, PPVs[-1], statline),
fontsize=8)
else:
fig.suptitle('%s (PDB: %s)\nPPV = %.2f \n%s' %
(c_filename, pdb_acc, PPVs[-1], statline),
fontsize=8)
else:
fig.suptitle('%s\nPPV = %.2f\n%s' %
(c_filename, PPVs[-1], statline),
fontsize=8)
#cmap = cm.get_cmap("binary")
#cmap.set_bad([1,1,1,0])
#contacts_np_masked = np.ma.array(contacts_np, mask=np.tri(contacts_np.shape[0], k=-1))
#sc = ax.imshow(contacts_np_masked.T, cmap=cmap)
#sc = ax.imshow(contacts_np, cmap=cmap)
#sc = ax.imshow(contacts_np + contacts_np.T, cmap=cm.binary, vmin=0.2, vmax=1.0, interpolation='none')
#sc = ax.scatter(contacts_x[::-1], contacts_y[::-1], marker='o', c=tp_colors[::-1], s=6, alpha=0.75, linewidths=0.0)
#sc = ax.scatter(contacts_y[::-1], contacts_x[::-1], marker='o', c=tp_colors[::-1], s=6, alpha=0.75, linewidths=0.0)
else:
#if c_filename.startswith('data'):
# acc = c_filename.split('/')[1]
#else:
# acc = c_filename.split('/')[-1]
fig.suptitle('%s\n%s' % (c_filename, statline), fontsize=8)
#sc = ax.imshow(contacts_np + contacts_np.T, cmap=cm.hot_r)
#sc = ax.imshow(contacts_np + contacts_np.T,
# cmap=cm.binary, vmin=th, vmax=1.0, interpolation='none')
img = get_colors(contacts_np, th=th, factor=factor)
sc = ax.imshow(img, interpolation='none')
#divider1 = make_axes_locatable(ax)
#cax1 = divider1.append_axes("right", size="2%", pad=0.05)
#plt.colorbar(sc, cax=cax1)
#plt.colorbar(sc, ax=ax)
#sc = ax.scatter(contacts_x[::-1], contacts_y[::-1],
# marker='o', c="black", s=6, alpha=0.75,
# linewidths=0.1, edgecolors='none')
#sc = ax.scatter(contacts_y[::-1], contacts_x[::-1], marker='o', c=scores[::-1], s=10, alpha=0.75, cmap=cm.hot_r, linewidths=0.1, edgecolors='none')
#plt.gca().set_xlim([0,ref_len])
#plt.gca().set_ylim([0,ref_len])
ax.grid()
ax.set_xlim([-unit, ref_len])
ax.set_ylim([-unit, ref_len])
#print ax.axis()
ax.axis([-unit, ref_len, -unit, ref_len])
#ax.invert_yaxis()
ax.set_autoscale_on(False)
if outfilename:
if outfilename.endswith('.pdf'):
pp = PdfPages(outfilename)
pp.savefig(fig)
pp.close()
ppstat = PdfPages(outfilename + "_statistics.pdf")
ppstat.savefig(statfig)
ppstat.close()
elif outfilename.endswith('.png'):
plt.savefig(outfilename)
else:
pp = PdfPages('%s.pdf' % outfilename)
pp.savefig(fig)
pp.close()
ppstat = PdfPages(outfilename + "_statistics.pdf")
ppstat.savefig(statfig)
ppstat.close()
else:
pp = PdfPages('%s_ContactMap.pdf' % c_filename)
pp.savefig(fig)
pp.close()
ppstat = PdfPages('%s_statistics.pdf' % c_filename)
ppstat.savefig(statfig)
ppstat.close()
def plot_map(acc,
fasta_filename,
contact_filename,
psipred_filename,
sep=',',
pdb_filename='',
chain='A'):
pdb_flag = pdb_filename.strip() != ''
rep_len = 1000
psipred_filename = '%s.horiz' % '.'.join(fasta_filename.split('.')[:-1])
ss = parse_psipred.horizontal(open(psipred_filename, 'r'))
ss_lst = get_ss_pos(ss)
print ss
print ss_lst
seq = parse_fasta.read_fasta(open(fasta_filename, 'r')).values()[0][0]
#ss = seq
ref_len = len(seq)
if pdb_flag:
#pdb_code = pdb_filename.split('/')[-1].split('.')[0]
res_lst = parse_pdb.get_coordinates(open(pdb_filename, 'r'), chain)
#cb_lst = parse_pdb.get_cb_coordinates(open(pdb_filename, 'r'), chain)
atom_seq = parse_pdb.get_atom_seq(open(pdb_filename, 'r'), chain)
#print seq
#print atom_seq
align = pairwise2.align.globalms(atom_seq, seq, 2, -1, -0.5, -0.1)
#print align[-1]
atom_seq_ali = align[-1][0]
seq_ali = align[-1][1]
print atom_seq_ali
print seq_ali
"""
# get 1D-domain assignments as char sequence
# 'D' = in domain / 'N' = not in domain
acc = fasta_filename.split('/')[-1].split('.')[0]
dom_seq, dom_lst = get_dom_seq(acc, ref_len, open('/bubo/home/h9/mircomic/glob/2013-05-29_human_repeats/query.txt', 'r'))
dom_seq_lst = map(int, list(dom_seq))
print dom_seq
"""
#dist_mat = calc_dist_matrix_heavy(ref_chain, ref_chain)
j = 0
gapped_res_lst = []
#gapped_cb_lst = []
for i in xrange(len(atom_seq_ali)):
if atom_seq_ali[i] == '-':
gapped_res_lst.append('-')
#gapped_cb_lst.append('-')
elif seq_ali[i] == '-':
j += 1
continue
else:
gapped_res_lst.append(res_lst[j])
#gapped_cb_lst.append(cb_lst[j])
j += 1
#print gapped_res_lst
print len(gapped_res_lst)
print len(res_lst)
#print len(gapped_cb_lst)
#print len(cb_lst)
print len(atom_seq)
print len(seq)
dist_mat = get_heavy_contacts(gapped_res_lst)
#dist_mat = get_cb_contacts(gapped_cb_lst)
heavy_cutoff = 5
cb_cutoff = 8
#ref_contact_map = (dist_mat < 8) & (dist_mat > 4)
ref_contact_map = dist_mat < heavy_cutoff
ref_contacts = np.where(dist_mat < heavy_cutoff)
#ref_contact_map = dist_mat < cb_cutoff
#ref_contacts = np.where(dist_mat < cb_cutoff)
ref_contacts_x = ref_contacts[0]
ref_contacts_y = ref_contacts[1]
"""
tmp_x = []
tmp_y = []
for i in range(len(ref_contacts_x)):
x = ref_contacts_x[i]
y = ref_contacts_y[i]
if y > x:
tmp_x.append(x)
tmp_y.append(y)
ref_contacts_x = tmp_x
ref_contacts_y = tmp_y
"""
#print dist_mat
#print ref_contact_map
#print 'ref_contacts=' + str(ref_contacts[1])
contacts = parse_contacts.parse(open(contact_filename, 'r'), sep)
#contacts_cut = contacts[0:ref_len]
contacts_x = []
contacts_y = []
scores = []
contact_dict = {}
count = 0
#for i in range(ref_len * 1):
for i in range(len(contacts)):
score = contacts[i][0]
c_x = contacts[i][1] - 1
c_y = contacts[i][2] - 1
pos_diff = abs(c_x - c_y)
too_close = pos_diff < 5
too_far = pos_diff > rep_len * 1.5
if not too_close:
contacts_x.append(c_x)
contacts_y.append(c_y)
scores.append(score)
count += 1
if count >= ref_len * 1.0:
break
if pdb_flag:
PPVs = []
inter_PPVs = []
intra_PPVs = []
TPs = []
inter_TPs = []
intra_TPs = []
FPs = []
inter_FPs = []
intra_FPs = []
for num_c in range(min(len(contacts_x), ref_len * 1))[1:]:
TP = 0.0
intra_TP = 0.0
inter_TP = 0.0
FP = 0.0
intra_FP = 0.0
inter_FP = 0.0
for i in range(num_c):
c_x = contacts_x[i]
c_y = contacts_y[i]
if atom_seq_ali[c_x] == '-':
continue
if atom_seq_ali[c_y] == '-':
continue
c_x_dom = in_dom(c_x, dom_lst)
c_y_dom = in_dom(c_y, dom_lst)
#print c_x_dom
#print c_y_dom
if ref_contact_map[c_x, c_y] > 0:
TP += 1.0
if c_x_dom != c_y_dom and c_x_dom != 0 and c_y_dom != 0:
inter_TP += 1.0
if c_x_dom == c_y_dom and c_x_dom != 0 and c_y_dom != 0:
intra_TP += 1.0
else:
FP += 1.0
if c_x_dom != c_y_dom and c_x_dom != 0 and c_y_dom != 0:
inter_FP += 1.0
if c_x_dom == c_y_dom and c_x_dom != 0 and c_y_dom != 0:
intra_FP += 1.0
#print '%s, %s, %s, %s, %s, %s' % (TP, FP, inter_TP, inter_FP, intra_TP, intra_FP)
if TP > 0 and FP > 0:
PPVs.append(TP / (TP + FP))
if inter_TP > 0 or inter_FP > 0:
inter_PPVs.append(inter_TP / (inter_TP + inter_FP))
if intra_TP > 0 or intra_FP > 0:
intra_PPVs.append(intra_TP / (intra_TP + intra_FP))
TPs.append(TP / ref_len)
FPs.append(FP / ref_len)
print len(PPVs)
if len(PPVs) > 0:
print PPVs[-1]
else:
PPVs.append(0.0)
print PPVs[-1]
if len(inter_PPVs) > 0:
print inter_PPVs[-1]
else:
inter_PPVs.append(0.0)
print inter_PPVs[-1]
if len(intra_PPVs) > 0:
print intra_PPVs[-1]
else:
intra_PPVs.append(0.0)
print intra_PPVs
#print TPs[-1]
#print FPs[-1]
PPV_file = open(
'/bubo/home/h9/mircomic/glob/2013-05-29_human_repeats/pconsc_predictions/PPV.txt',
'a')
PPV_file.write('%s\t%s\n' % (acc, PPVs[-1]))
PPV_file.close()
intra_PPV_file = open(
'/bubo/home/h9/mircomic/glob/2013-05-29_human_repeats/pconsc_predictions/intra_PPV.txt',
'a')
intra_PPV_file.write('%s\t%s\n' % (acc, intra_PPVs[-1]))
intra_PPV_file.close()
inter_PPV_file = open(
'/bubo/home/h9/mircomic/glob/2013-05-29_human_repeats/pconsc_predictions/inter_PPV.txt',
'a')
inter_PPV_file.write('%s\t%s\n' % (acc, inter_PPVs[-1]))
inter_PPV_file.close()
### get pairwise distances from the alignment
#alinum = parse_jones.get_numeric('%s.jones' % '.'.join(contact_filename.split('.')[:-1]))
#alidist = squareform(pdist(alinum.T, 'euclidean'))
#alidistlog = np.log(alidist)
"""
dot_matrix = dotter.calc_dot_matrix(seq)
tmp_dot_matrix = dot_matrix
for (i,j), score in np.ndenumerate(dot_matrix):
if i > j:
tmp_dot_matrix[i, j] = 0.0
dot_matrix = tmp_dot_matrix
"""
fig = plt.figure()
ax = fig.add_subplot(111)
for i in range(len(ss)):
if ss[i] == 'H':
plt.plot(i, i, 'o', c='#8B0043', mec="#8B0043", markersize=2)
#plt.plot(i, i, 'o', c='#999999', mec="#444444")
if ss[i] == 'E':
plt.plot(i, i, 'D', c='#0080AD', mec="#0080AD", markersize=2)
#plt.plot(i, i, 'D', c='#999999', mec="#444444", markersize=6)
if ss[i] == 'C':
continue
#plt.plot(i, i, 'D', c='#999999', mec='#999999', markersize=3)
#plt.plot(i, i, 'D', c='#999999', mec='#999999', markersize=3)
"""
for dom in dom_lst:
start = dom[0] - 1
end = dom[1] - 1
dom_len = end - start
print '%d - %d' % (start, end)
ax.add_patch(plt.Rectangle((start, start), dom_len, dom_len, facecolor='#EEEEEE', edgecolor='black', lw=0.5, zorder=0))
for dom2 in dom_lst:
if dom2 == dom:
continue
start2 = dom2[0] - 1
end2 = dom2[1] - 1
dom_len2 = end2 - start2
ax.add_patch(plt.Rectangle((start, start2), dom_len, dom_len2, facecolor='#EEEEEE', edgecolor='black', lw=0.5, zorder=0, alpha=0.3))
"""
ss_col_dict = {'H': '#8B0043', 'E': '#0080AD', 'HE': '#3F467D'}
for ss_elem in ss_lst:
start = ss_elem[0] - 1
end = ss_elem[1] - 1
elem_len = end - start
curr_ss = ss[start]
#print '%d - %d' % (start, end)
ax.add_patch(
plt.Rectangle((start, start),
elem_len,
elem_len,
facecolor=ss_col_dict[curr_ss],
edgecolor='black',
lw=0.5,
zorder=0,
alpha=0.5))
for ss_elem2 in ss_lst:
if ss_elem2 == ss_elem:
continue
start2 = ss_elem2[0] - 1
end2 = ss_elem2[1] - 1
elem_len2 = end2 - start2
curr_ss2 = ss[start2]
if curr_ss != curr_ss2: # interaction between sheet and helix
curr_ss2 = 'HE'
ax.add_patch(
plt.Rectangle((start, start2),
elem_len,
elem_len2,
facecolor=ss_col_dict[curr_ss2],
edgecolor='black',
lw=0.5,
zorder=0,
alpha=0.2))
#ax.imshow(dot_matrix, origin='lower', cmap=cm.binary)
if pdb_flag:
ax.scatter(ref_contacts_x,
ref_contacts_y,
marker='o',
c='#CCCCCC',
lw=0)
#ax.scatter(range(ref_len), range(ref_len), marker='d', c=dom_seq_lst, lw=0, edgecolor=dom_seq_lst, cmap=cm.spectral_r)
#plt.plot(ref_contacts_x, ref_contacts_y, 'o', c='#CCCCCC', mec='#CCCCCC')
fig.suptitle(
'%s\nPPV = %.2f intra-PPV = %.2f inter-PPV = %.2f' %
(contact_filename, PPVs[-1], intra_PPVs[-1], inter_PPVs[-1]))
else:
fig.suptitle('%s - %s' % (acc, contact_filename))
sc = ax.scatter(contacts_x[::-1],
contacts_y[::-1],
marker='o',
c=scores[::-1],
s=4,
alpha=0.75,
cmap=cm.jet,
linewidths=0.5)
#sc = ax.scatter(contacts_nf_y[::-1], contacts_nf_x[::-1], marker='o', c=scores_nf[::-1], s=8, alpha=0.75, cmap=cm.jet, linewidths=0.5)
plt.gca().set_xlim([0, ref_len])
plt.gca().set_ylim([0, ref_len])
plt.colorbar(sc)
#cbar = plt.colorbar(ax, ticks=[min(scores), max(scores)])
#cbar.ax.set_yticklabels(['Low', 'High'])
#cbar.set_label(r'Contact Score')
pp = PdfPages('%s_ContactMap.pdf' % contact_filename)
pp.savefig(fig)
pp.close()
#outfile = open('%s.contacts' % '.'.join(contact_filename.split('.')[0:-1]),'w')
#for i in range(len(scores)):
# outfile.write('%s,%s,%s\n' % (int(contacts_x[i] + 1), int(contacts_y[i] + 1), scores[i]))
if pdb_flag:
fig.clf()
ax2 = fig.add_subplot(111)
ax2.plot(PPVs)
pp = PdfPages('%s_PPVs.pdf' % contact_filename)
pp.savefig(fig)
pp.close()
def plot_map(fasta_filename,
c_filename,
factor=1.0,
th=-1,
f_obs=-1,
c2_filename='',
psipred_horiz_fname='',
psipred_vert_fname='',
pdb_filename='',
is_heavy=False,
chain='',
sep=',',
outfilename='',
ali_filename='',
meff_filename='',
name='',
start=0,
end=-1,
pdb_start=0,
pdb_end=-1,
noalign=False,
pdb_alignment='',
pdb_id='',
binary=False):
#acc = c_filename.split('.')[0]
#acc = fasta_filename.split('.')[0][:4]
if name == '':
acc = '.'.join(os.path.basename(fasta_filename).split('.')[:-1])
else:
acc = name
### get sequence
seq = parse_fasta.read_fasta(open(fasta_filename, 'r')).values()[0][0]
ref_len = len(seq)
### trim sequence according to given positions
### default: take full sequence
if end == -1:
end = ref_len
seq = seq[start:end]
ref_len = len(seq)
unit = (ref_len / 50.0)
### get top "factor" * "ref_len" predicted contacts
contacts = parse_contacts.parse(open(c_filename, 'r'), sep)
contacts_np = parse_contacts.get_numpy_cmap(contacts, seq_len=ref_len)
contacts_np = contacts_np[start:end, start:end]
contacts_x = []
contacts_y = []
scores = []
contact_dict = {}
count = 0
for i in range(len(contacts)):
score = contacts[i][0]
c_x = contacts[i][1] - 1
c_y = contacts[i][2] - 1
# only look at contacts within given range
# default: take full sequence range into account
if c_x < start or c_x >= end:
continue
if c_y < start or c_y >= end:
continue
pos_diff = abs(c_x - c_y)
too_close = pos_diff < 5
if not too_close:
contacts_x.append(c_x - start)
contacts_y.append(c_y - start)
scores.append(score)
count += 1
if count >= ref_len * factor and th == -1 and f_obs == -1:
th = score
break
if score < th and not th == -1 and f_obs == -1:
factor = count / float(ref_len)
break
# if cutoff by fraction of observed contacts:
# take all contacts and cut list after reading pdb
#if f_obs != -1:
# factor = ref_len
# th = -1
### start plotting
fig = plt.figure(figsize=(8, 8), dpi=96, facecolor='w')
ax = fig.add_subplot(111) #, aspect='auto')
ax.set_adjustable('box-forced')
ax.tick_params(direction='out', right='off', top='off')
ax.set_xlim([-unit, ref_len])
ax.set_ylim([-unit, ref_len])
### plot alignment coverage if alignemnt given
if ali_filename or meff_filename:
# adjust overall canvas
ax = plt.subplot2grid((8, 8), (1, 1), colspan=7,
rowspan=7) #, aspect='auto')
#ax.set_adjustable('box-forced')
#ax.set_autoscale_on(False)
ax.autoscale(False)
ax.tick_params(direction='out',
labelleft='off',
right='off',
top='off')
ax.set_xlim([-unit, ref_len])
ax.set_ylim([-unit, ref_len])
if ali_filename:
coverage_lst, M = get_ali_coverage(ali_filename)
max_cover = M
elif meff_filename:
coverage_lst, Meff = get_meff_coverage(meff_filename)
max_cover = Meff
#max_cover = max(coverage_lst)
coverage_lst = coverage_lst[start:end]
#lt = pow(10, max(1,floor(log10(max_cover)) - 1))
#upper = int(ceil(max_cover/float(lt)) * lt)
ax2 = plt.subplot2grid((8, 8), (1, 0), rowspan=7, sharey=ax)
#ax2.set_adjustable('box-forced')
#ax2.set_autoscale_on(False)
ax2.autoscale(False)
#print len([0]+coverage_lst+[0])
#print len([0]+range(ref_len)+[ref_len-1])
ax2.plot([0] + coverage_lst + [0],
[0] + range(ref_len) + [ref_len - 1],
'k',
lw=0)
ax2.axvline(x=max_cover * 0.25, lw=0.5, c='black', ls=':')
ax2.axvline(x=max_cover * 0.5, lw=0.5, c='black', ls=':')
ax2.axvline(x=max_cover * 0.75, lw=0.5, c='black', ls=':')
ax2.fill([0] + coverage_lst + [0],
[0] + range(ref_len) + [ref_len - 1],
facecolor='gray',
lw=0,
alpha=0.5)
ax2.set_xticks([0, max_cover])
ax2.tick_params(axis='x', top='off', direction='out')
ax2.invert_xaxis()
#ax2.spines['top'].set_visible(False)
#ax2.spines['left'].set_visible(False)
#ax.get_xaxis().tick_bottom()
#ax.get_yaxis().tick_right()
ax2.grid()
ax2.set_ylim([-unit, ref_len])
ax3 = plt.subplot2grid((8, 8), (0, 1), colspan=7, sharex=ax)
#ax3.set_adjustable('box-forced')
#ax3.set_autoscale_on(False)
ax3.autoscale(False)
ax3.plot([0] + range(ref_len) + [ref_len - 1],
[0] + coverage_lst + [0],
'k',
lw=0)
ax3.axhline(y=max_cover * 0.25, lw=0.5, c='black', ls=':')
ax3.axhline(y=max_cover * 0.5, lw=0.5, c='black', ls=':')
ax3.axhline(y=max_cover * 0.75, lw=0.5, c='black', ls=':')
ax3.fill([0] + range(ref_len) + [ref_len - 1],
[0] + coverage_lst + [0],
facecolor='gray',
lw=0,
alpha=0.5)
#ax3.xaxis.tick_top()
ax3.set_yticks([0, max_cover])
ax3.tick_params(labelbottom='off')
ax2.tick_params(axis='y', right='off', direction='out', left='on')
#ax3.spines['top'].set_visible(False)
#ax3.spines['right'].set_visible(False)
#ax.get_xaxis().tick_top()
#ax.get_yaxis().tick_left()
ax3.grid()
ax3.set_xlim([-unit, ref_len])
### plot secondary structure along axis if given
if psipred_horiz_fname or psipred_vert_fname:
if psipred_horiz_fname:
ss = parse_psipred.horizontal(open(psipred_horiz_fname, 'r'))
else:
ss = parse_psipred.vertical(open(psipred_vert_fname, 'r'))
ss = ss[start:end]
assert len(ss) == ref_len
ax.axhline(y=0, lw=1, c='black')
ax.axvline(x=0, lw=1, c='black')
for i in range(len(ss)):
if ss[i] == 'H':
#ax.plot(-unit/2, i, 's', c='#8B0043', mec="#8B0043")#, markersize=2)
#ax.plot(i, -unit/2, 's', c='#8B0043', mec="#8B0043")#, markersize=2)
#ax.plot(i, -unit/2, 's', c='#8B0043', mec="#8B0043")#, markersize=2)
ax.add_patch(
plt.Rectangle((-unit, i - 0.5),
unit,
1,
edgecolor='#8B0043',
facecolor="#8B0043"))
ax.add_patch(
plt.Rectangle((i - 0.5, -unit),
1,
unit,
edgecolor='#8B0043',
facecolor="#8B0043"))
if ss[i] == 'E':
ax.add_patch(
plt.Rectangle((-unit, i - 0.5),
unit,
1,
edgecolor='#0080AD',
facecolor="#0080AD"))
ax.add_patch(
plt.Rectangle((i - 0.5, -unit),
1,
unit,
edgecolor='#0080AD',
facecolor="#0080AD"))
#ax.plot(-unit/2, i, 's', c='#0080AD', mec="#0080AD")#, markersize=2)
#ax.plot(i, -unit/2, 's', c='#0080AD', mec="#0080AD")#, markersize=2)
if ss[i] == 'C':
continue
### plot reference contacts in the background if given
if pdb_filename:
res_lst = parse_pdb.get_coordinates(open(pdb_filename, 'r'), chain)
cb_lst = parse_pdb.get_cb_coordinates(open(pdb_filename, 'r'), chain)
atom_seq = parse_pdb.get_atom_seq(open(pdb_filename, 'r'), chain)
# if not true there is some serious problem with the provided pdb file
assert len(res_lst) == len(cb_lst) == len(atom_seq)
### trim PDB sequence according to given positions
### default: take full sequence
if pdb_end == -1:
pdb_end = len(res_lst)
res_lst = res_lst[pdb_start:pdb_end]
cb_lst = cb_lst[pdb_start:pdb_end]
atom_seq = atom_seq[pdb_start:pdb_end]
#print atom_seq
#print seq
if noalign:
dist_mat = get_cb_contacts(cb_lst)
cb_cutoff = 8
ref_contact_map = dist_mat < cb_cutoff
PPV, TP, FP = get_ppvs(contacts_x, contacts_y, ref_contact_map,
ref_len, factor)
tp_colors = get_tp_colors(contacts_x, contacts_y, ref_contact_map)
else:
if pdb_alignment and pdb_id:
#align = parse_hhblits_hhr.parse_alignments(pdb_alignment)[pdb_id]
#atom_seq_ali = align[0][0]
#seq_ali = align[0][1]
seq_ali, atom_seq_ali = parse_a3m.get_pairwise(
pdb_alignment, pdb_id)
seqres_seq = atom_seq_ali.replace('-', '')
#print seqres_seq
#print atom_seq_ali
#print seq_ali
#print ""
align_seqres = pairwise2.align.globalms(
atom_seq, seqres_seq, 2, -1, -0.5, -0.1)
atom_seq_ali0 = align_seqres[-1][0]
seqres_seq_ali0 = align_seqres[-1][1]
#print ""
#print atom_seq_ali0
#print seqres_seq_ali0
#print ""
atom_seq_ali, seq_ali = embedd_alignment(
atom_seq_ali0, seqres_seq_ali0, atom_seq_ali, seq_ali)
#print atom_seq_ali1
#print seq_ali1
#print ""
else:
matrix = matlist.blosum62
#align = pairwise2.align.globalms(atom_seq, seq, 2, -1, -0.5, -0.1)
#align = pairwise2.align.localds(atom_seq, seq, matrix, -11, -1)
align = pairwise2.align.globalds(atom_seq, seq, matrix, -25,
-1)
atom_seq_ali = align[-1][0]
seq_ali = align[-1][1]
#print atom_seq_ali
#print seq_ali
#print len(atom_seq), len(seq), len(res_lst), len(cb_lst)
#print len(atom_seq_ali), len(seq_ali)
j = 0
gapped_res_lst = []
gapped_cb_lst = []
for i in xrange(len(atom_seq_ali)):
if atom_seq_ali[i] == '-':
if seq_ali[i] == '-':
continue
gapped_res_lst.append('-')
gapped_cb_lst.append('-')
elif seq_ali[i] == '-':
j += 1
else:
gapped_res_lst.append(res_lst[j])
gapped_cb_lst.append(cb_lst[j])
j += 1
if is_heavy:
dist_mat = get_heavy_contacts(gapped_res_lst)
heavy_cutoff = 5
ref_contact_map = dist_mat < heavy_cutoff
ref_contacts = np.where(dist_mat < heavy_cutoff)
else:
dist_mat = get_cb_contacts(gapped_cb_lst)
cb_cutoff = 8
ref_contact_map = dist_mat < cb_cutoff
ref_contacts = np.where(dist_mat < cb_cutoff)
#ref_contacts = np.where(np.ma.array(dist_mat, mask=np.tri(dist_mat.shape[0]), fill_value=float("inf")) < cb_cutoff)
ref_contacts_x = ref_contacts[0]
ref_contacts_y = ref_contacts[1]
# if f_obs given, take top f_obs * num_obs contacts:
if f_obs != -1:
num_obs = sum(ref_contacts_x - ref_contacts_y >= 5)
num_top = int(ceil(f_obs * num_obs))
contacts_x = contacts_x[:num_top]
contacts_y = contacts_y[:num_top]
scores = scores[:num_top]
th_obs = scores[-1]
else:
th_obs = th
PPVs, TPs, FPs = get_ppvs(contacts_x,
contacts_y,
ref_contact_map,
ref_len,
factor,
atom_seq_ali=atom_seq_ali)
tp_colors = get_tp_colors(contacts_x,
contacts_y,
ref_contact_map,
atom_seq_ali=atom_seq_ali)
if not c2_filename:
img = get_colors(contacts_np,
ref_contact_map=dist_mat,
th=th_obs,
binary=binary)
sc = ax.imshow(img, interpolation='none')
else:
# plot native contacts in background
img = get_ref_img(dist_mat)
sc = ax.imshow(img, interpolation='none')
print '%s %s %s %s' % (acc, PPVs[-1], TPs[-1], FPs[-1])
cmap = cm.get_cmap("binary")
cmap.set_bad([1, 1, 1, 0])
dist_mat_masked = np.ma.array(dist_mat,
mask=np.tri(dist_mat.shape[0], k=-1))
#sc = ax.imshow(s_score_vec(dist_mat_masked, 5), cmap=cmap, interpolation='none')
ref_contacts_diag_x = []
ref_contacts_diag_y = []
for i in range(len(ref_contacts_x)):
x_i = ref_contacts_x[i]
y_i = ref_contacts_y[i]
if not dist_mat_masked.mask[x_i, y_i] and abs(x_i - y_i) >= 5:
ref_contacts_diag_x.append(x_i)
ref_contacts_diag_y.append(y_i)
#ax.scatter(ref_contacts_diag_x, ref_contacts_diag_y, marker='+', c='#000000')
### plot predicted contacts from second contact map if given
if c2_filename:
contacts2 = parse_contacts.parse(open(c2_filename, 'r'))
contacts2_x = []
contacts2_y = []
scores2 = []
contact_dict2 = {}
count = 0
for i in range(len(contacts2)):
score = contacts2[i][0]
c_x = contacts2[i][1] - 1
c_y = contacts2[i][2] - 1
pos_diff = abs(c_x - c_y)
too_close = pos_diff < 5
if not too_close:
contacts2_x.append(c_x - start)
contacts2_y.append(c_y - start)
scores2.append(score)
count += 1
if count >= ref_len * factor and th == -1 and f_obs == -1:
th = score
break
if score < th and not th == -1 and f_obs == -1:
factor = count / float(ref_len)
break
# if cutoff by fraction of observed contacts:
# stop at num_top = f_obs * num_obs
if pdb_filename and f_obs != -1:
if count >= num_top:
factor = count / float(ref_len)
th = score
break
### use TP/FP color coding if reference contacts given
if pdb_filename:
#PPVs2, TPs2, FPs2 = get_ppvs(contacts2_x, contacts2_y, ref_contact_map, atom_seq_ali, ref_len, factor)
tp2_colors = get_tp_colors(contacts2_x, contacts2_y,
ref_contact_map, atom_seq_ali)
#print '%s %s %s %s' % (acc, PPVs2[-1], TPs2[-1], FPs2[-1])
#fig.suptitle('%s\nPPV (upper left) = %.2f | PPV (lower right) = %.2f' % (acc, PPVs[-1], PPVs2[-1]))
sc = ax.scatter(contacts2_y[::-1],
contacts2_x[::-1],
marker='s',
c=tp2_colors[::-1],
s=4,
alpha=1,
lw=0,
edgecolor=tp2_colors[::-1])
sc = ax.scatter(contacts_x[::-1],
contacts_y[::-1],
marker='s',
c=tp_colors[::-1],
s=4,
alpha=1,
lw=0,
edgecolor=tp_colors[::-1])
else:
sc = ax.scatter(contacts2_y[::-1],
contacts2_x[::-1],
marker='0',
c='#D70909',
edgecolor='#D70909',
s=6,
linewidths=0.5)
sc = ax.scatter(contacts_x[::-1],
contacts_y[::-1],
marker='o',
c='#004F9D',
edgecolor='#004F9D',
s=6,
linewidths=0.5)
### plot predicted contacts from first contact map on both triangles
### if no second contact map given
else:
if pdb_filename:
pdb_acc = parse_pdb.get_acc(open(pdb_filename))
if pdb_acc:
if chain:
fig.suptitle('%s (PDB: %s, chain %s)\nPPV = %.2f' %
(acc, pdb_acc, chain, PPVs[-1]))
else:
fig.suptitle('%s (PDB: %s)\nPPV = %.2f' %
(acc, pdb_acc, PPVs[-1]))
else:
fig.suptitle('%s\nPPV = %.2f' % (acc, PPVs[-1]))
#cmap = cm.get_cmap("binary")
#cmap.set_bad([1,1,1,0])
#contacts_np_masked = np.ma.array(contacts_np, mask=np.tri(contacts_np.shape[0], k=-1))
#sc = ax.imshow(contacts_np_masked.T, cmap=cmap)
#sc = ax.imshow(contacts_np, cmap=cmap)
#sc = ax.imshow(contacts_np + contacts_np.T, cmap=cm.binary, vmin=0.2, vmax=1.0, interpolation='none')
#sc = ax.scatter(contacts_x[::-1], contacts_y[::-1], marker='o', c=tp_colors[::-1], s=6, alpha=0.75, linewidths=0.0)
#sc = ax.scatter(contacts_y[::-1], contacts_x[::-1], marker='o', c=tp_colors[::-1], s=6, alpha=0.75, linewidths=0.0)
else:
#if c_filename.startswith('data'):
# acc = c_filename.split('/')[1]
#else:
# acc = c_filename.split('/')[-1]
fig.suptitle('%s' % acc)
#sc = ax.imshow(contacts_np + contacts_np.T, cmap=cm.hot_r)
#sc = ax.imshow(contacts_np + contacts_np.T,
# cmap=cm.binary, vmin=th, vmax=1.0, interpolation='none')
img = get_colors(contacts_np, th=th)
sc = ax.imshow(img, interpolation='none')
#divider1 = make_axes_locatable(ax)
#cax1 = divider1.append_axes("right", size="2%", pad=0.05)
#plt.colorbar(sc, cax=cax1)
#plt.colorbar(sc, ax=ax)
#sc = ax.scatter(contacts_x[::-1], contacts_y[::-1],
# marker='o', c="black", s=6, alpha=0.75,
# linewidths=0.1, edgecolors='none')
#sc = ax.scatter(contacts_y[::-1], contacts_x[::-1], marker='o', c=scores[::-1], s=4, alpha=0.75, cmap=cm.hot_r, linewidths=0.1, edgecolors='none')
#plt.gca().set_xlim([0,ref_len])
#plt.gca().set_ylim([0,ref_len])
ax.grid()
ax.set_xlim([-unit, ref_len])
ax.set_ylim([-unit, ref_len])
#print ax.axis()
ax.axis([-unit, ref_len, -unit, ref_len])
#ax.invert_yaxis()
ax.set_autoscale_on(False)
if outfilename:
if outfilename.endswith('.pdf'):
pp = PdfPages(outfilename)
pp.savefig(fig)
pp.close()
elif outfilename.endswith('.eps'):
plt.savefig(outfilename, format='eps', dpi=300)
elif outfilename.endswith('.png'):
plt.savefig(outfilename)
else:
pp = PdfPages('%s.pdf' % outfilename)
pp.savefig(fig)
pp.close()
else:
pp = PdfPages('%s_ContactMap.pdf' % c_filename)
pp.savefig(fig)
pp.close()
return c_filt
def cysteine_filter(c_lst, seq):
"""Filters contacts from cysteines
@param c_lst contact list (as given by parsing/parse_contacts.py)
@param seq string of one-letter coded amino acid sequence
Ensures: len(c_lst) == len(c_filt), only contact weights are changed
@return [(score, residue a, residue b)]
"""
c_filt = []
return c_filt
if __name__ == "__main__":
c_filename = sys.argv[1]
psipred_filename = sys.argv[2]
cfilt_filename = sys.argv[3]
#seq_filename = sys.argv[3]
c_lst = parse_contacts.parse(open(c_filename, 'r'))
ss_seq = parse_psipred.horizontal(open(psipred_filename, 'r'))
c_filt = secstruct_filter(c_lst, ss_seq)
cfilt_file = open(cfilt_filename, 'w')
parse_contacts.write(c_filt, cfilt_file)
cfilt_file.close()