def visualize_nucl_substitution_matrix(shms_df, output_fname, log):
nucl_list = ['A', 'C', 'G', 'T']
nucl_matrix = []
for n in nucl_list:
nucl_matrix.append([0] * len(nucl_list))
num_shms = 0
for it in shms_df:
read_shms = shms_df[it]
for shm in read_shms:
if not shm.is_substitution():
continue
if nucl_is_valid(shm.read_nucl) and nucl_is_valid(shm.gene_nucl):
nucl_matrix[nucl_list.index(shm.read_nucl)][nucl_list.index(
shm.gene_nucl)] += 1
num_shms += 1
for i in range(0, len(nucl_matrix)):
for j in range(0, len(nucl_matrix[i])):
nucl_matrix[i][j] = float(nucl_matrix[i][j]) / float(num_shms)
fig, ax = plt.subplots()
sns.heatmap(nucl_matrix,
cmap=plt.cm.Blues,
xticklabels=nucl_list,
yticklabels=nucl_list,
square=True,
ax=ax)
ax.tick_params(labelsize=14)
plt.xticks(fontsize=12)
plt.yticks(fontsize=12, rotation='horizontal')
plt.xlabel("To", fontsize=14)
plt.ylabel("From", fontsize=14, rotation='horizontal')
utils.output_figure(output_fname, "Nucleotide substitution heatmap", log)
def visualize_largest_group_aa_variability(labeling_df, region, region_name,
output_fname, log):
region_seq = list(labeling_df[region])
max_group = get_region_largest_group(region_seq)
if len(max_group) == 0:
return
group_len = len(max_group[0])
if group_len % 3 != 0:
print("Largest " + region_name + " is not out-of-frame")
return
aa_len = group_len // 3
aa_seqs = [Seq(cdr).translate(to_stop=True) for cdr in max_group]
aa_dict = {'Position': [], 'Hidrophobicity': []}
for aa_seq in aa_seqs:
aa_row = [utils.hydrophoby_dict[aa] for aa in aa_seq]
for i in range(len(aa_row)):
aa_dict['Position'].append(i + 1)
aa_dict['Hidrophobicity'].append(aa_row[i])
plt.figure()
sns.barplot(x='Position',
y='Hidrophobicity',
data=aa_dict,
order=range(1, aa_len + 1),
color='blue')
plt.xlabel('Position (aa)', fontsize=14)
plt.ylabel('Hidrophobicity', fontsize=14)
plt.ylim(
min(utils.hydrophoby_dict.values()) - 10,
max(utils.hydrophoby_dict.values()) + 10)
utils.output_figure(output_fname, region_name + " aa variability", log)
def visualize_largest_region_nucls(labeling_df, region, region_name,
output_fname, log):
region_seq = list(labeling_df[region])
max_group = get_region_largest_group(region_seq)
if len(max_group) == 0:
return
nucl_dict = get_nucls_lists(max_group)
x = np.array(range(0, len(max_group[0])))
x_l = [str(i) for i in range(1, len(max_group[0]) + 1)]
acgt = nucl_dict['A'] + nucl_dict['C'] + nucl_dict['G'] + nucl_dict['T']
cgt = nucl_dict['C'] + nucl_dict['G'] + nucl_dict['T']
gt = nucl_dict['G'] + nucl_dict['T']
#sns.set_color_codes("pastel")
sns.set_color_codes("muted")
f, ax = plt.subplots(figsize=(15, 6))
sns.barplot(x=x, y=acgt, label="A", color='b')
sns.barplot(x=x, y=cgt, label="C", color='g')
sns.barplot(x=x, y=gt, label="G", color='r')
sns.barplot(x=x, y=nucl_dict['T'], label="T", color='orange')
plt.ylim(0, 115)
ax.legend(ncol=4, loc="upper center", frameon=True, fontsize=16)
plt.xlabel(region_name + ' position (nt)', fontsize=16)
plt.ylabel('Nucleotide %', fontsize=16)
plt.xticks(x, x_l, fontsize=14)
plt.yticks(fontsize=14)
utils.output_figure(output_fname, region_name + " nucleotide distribution",
log)
def output_shm_stats_for_isotype(num_shms, shm_pos, isotype, output_prefix,
log):
plt.figure(1)
plt.subplot(211)
# plot for SHM positions
plt.hist(shm_pos, color=isotype_colors[isotype], alpha=.75, bins=50)
#cdr_color = "#EFBEBE"
#plt.gca().add_patch(patches.Rectangle((cdr_positions[isotype]['CDR1'][0], 0),
# cdr_positions[isotype]['CDR1'][1] - cdr_positions[isotype]['CDR1'][0],
# max(n) + 2, facecolor= cdr_color, lw = 0))
#plt.gca().add_patch(patches.Rectangle((cdr_positions[isotype]['CDR2'][0], 0),
# cdr_positions[isotype]['CDR2'][1] - cdr_positions[isotype]['CDR2'][0],
# max(n) + 2, facecolor= cdr_color, lw = 0))
#plt.gca().add_patch(patches.Rectangle((cdr_positions[isotype]['CDR3'][0], 0),
# cdr_positions[isotype]['CDR3'][1] - cdr_positions[isotype]['CDR3'][0],
# max(n) + 2, facecolor= cdr_color, lw = 0))
#n, bins, p = pylab.hist(shm_pos, color = isotype_colors[isotype], bins = 50)
plt.xlabel("Relative position of " + isotype + "V SHM in read",
fontsize=16)
plt.ylabel("# SHMs", fontsize=16)
plt.xticks(fontsize=14)
plt.yticks(fontsize=14)
plt.title("SHMs in " + isotype + "V", fontsize=18)
# plot for SHM number
plt.subplot(212)
plt.hist(num_shms, color=isotype_colors[isotype], bins=50, alpha=.75)
plt.xlabel("# SHMs in " + isotype + "V", fontsize=16)
plt.ylabel("# sequences", fontsize=16)
plt.xticks(fontsize=14)
plt.yticks(fontsize=14)
output_fname = output_prefix + "_" + isotype + "V"
utils.output_figure(output_fname,
"Distribution of # SHMs in " + isotype + "V segments",
log)
def visualize_length_abundance_dist(labeling_df, region, region_name,
output_fname, log):
region_seq = list(labeling_df[region])
region_dict = dict()
for seq in region_seq:
if seq not in region_dict:
region_dict[seq] = 0
region_dict[seq] += 1
abun = [] #np.array()
lens = [] #np.array()
for seq in region_dict:
if region_dict[seq] == 1:
continue
abun.append(region_dict[seq])
lens.append(len(seq))
abun = np.asarray(abun)
lens = np.asarray(lens)
f, ax = plt.subplots()
sns.jointplot(abun, lens, size=6)
#plt.xlabel(region_name + ' abundance', fontsize = 14)
#ax.xaxis.set_label_position('top')
#plt.ylabel(region_name + ' length', fontsize = 14)
#plt.xticks(fontsize = 14)
#plt.yticks(fontsize = 14)
#plt.xlim(-1, abun.max() + 1)
#plt.ylim(-1, lens.max() + 1)
utils.output_figure(
output_fname,
region_name + " joint distribution of abundances & lengths", log)
def visualize_indel_shm_lengths(shm_df, output_fname, log):
prev_read_pos = -1
prev_gene_pos = -1
insertion_length = []
deletions_lengths = []
in_len = 0
del_len = 0
for it in shm_df:
read_shms = shm_df[it]
for shm in read_shms:
if shm.is_deletion():
if shm.gene_pos - prev_gene_pos == 1:
del_len += 1
else:
if del_len > 0:
deletions_lengths.append(del_len)
del_len = 1
prev_gene_pos = shm.gene_pos
if shm.is_insertion():
if shm.read_pos - prev_read_pos == 1:
in_len += 1
else:
if in_len > 0:
insertion_length.append(in_len)
in_len = 1
prev_read_pos = shm.read_pos
if in_len != 0:
insertion_length.append(in_len)
if del_len != 0:
deletions_lengths.append(del_len)
dt = []
labels = []
max_x_value = 0
if len(deletions_lengths) > 10:
dt.append(deletions_lengths)
labels.append("Deletions")
if len(insertion_length) > 10:
dt.append(insertion_length)
labels.append("Insertions")
if len(dt) == 0:
log.info(
"Output contains very low number of indel SHMs. Plot drawing was skipped"
)
return
plt.hist(dt, label=labels, bins=50)
plt.legend(loc='upper center', ncol=len(dt), fontsize=14)
plt.xlabel("Insertion / deletion SHM length", fontsize=16)
plt.ylabel("# insertion / deletion SHMs", fontsize=16)
xlim_right = 0
if len(deletions_lengths) != 0:
xlim_right = max(deletions_lengths)
if len(insertion_length) != 0:
xlim_right = max(xlim_right, max(insertion_length))
plt.xlim(.5, xlim_right + .5)
plt.xticks(range(0, xlim_right + 1), fontsize=14)
plt.yticks(fontsize=14)
utils.output_figure(output_fname,
"Distribution of insertion/deletion SHM lengths", log)
def visualize_special_shm_positions(shm_df, syn_output_fname,
special_output_fname, log):
synonymous_pos = []
stop_codon_pos = []
deletion_pos = []
insertion_pos = []
for it in shm_df:
read_shms = shm_df[it]
for shm in read_shms:
if not it.is_variable():
continue
relative_pos = float(shm.read_pos) / float(it.read_len)
if shm.synonymous:
synonymous_pos.append(relative_pos)
elif shm.to_stop_codon:
stop_codon_pos.append(relative_pos)
elif shm.is_deletion():
deletion_pos.append(relative_pos)
elif shm.is_insertion():
insertion_pos.append(relative_pos)
output_synonymous_shms(synonymous_pos, syn_output_fname, log)
pos = []
labels = []
colors = []
plt.figure(figsize=(12, 9))
#sns.distplot(synonymous_pos, hist = False, label = "Synonymous SHMs", color = 'r')
#if len(stop_codon_pos) > 100:
# pos.append(stop_codon_pos)
# labels.append('Stop codon')
# colors.append('g')
# #sns.distplot(stop_codon_pos, hist = False, label = "Stop codon SHMs", color = 'g')
if len(deletion_pos) > 10:
pos.append(deletion_pos)
labels.append('Deletions')
colors.append('b')
#sns.distplot(deletion_pos, hist = False, label = "Deletion SHMs", color = 'b')
if len(insertion_pos) > 10:
pos.append(insertion_pos)
labels.append('Insertions')
colors.append('g')
if len(pos) == 0:
log.info(
"Output contains very low number of special SHMs. Plot drawing will be skipped"
)
return
#sns.distplot(insertion_pos, hist = False, label = "Insertion SHMs", color = 'orange')
plt.hist(pos, color=colors, label=labels, bins=100 / len(pos))
plt.xlim(0, .75)
plt.legend(loc='upper center',
ncol=len(pos),
fontsize=12,
bbox_to_anchor=(0.5, -0.07))
plt.xlabel("Relative position of V SHM in read", fontsize=14)
plt.ylabel("# SHMs", fontsize=14)
plt.xticks(fontsize=12)
plt.yticks(fontsize=12)
utils.output_figure(special_output_fname,
"Distribution of indel V SHM positions in read", log)
def visualize_region_lengths(labeling_df, region, region_name, output_fname,
log):
region_seq = list(labeling_df[region])
region_len = [len(s) for s in region_seq if len(s) > 1]
plt.figure()
plt.hist(region_len)
plt.xlabel(region_name + ' length (nt)', fontsize=14)
plt.ylabel('# ' + region_name + 's', fontsize=14)
utils.output_figure(output_fname, region_name + " length distribution",
log)
def output_shm_stats_for_isotype(num_shms, locus, output_fname, log):
plt.figure()
plt.hist(num_shms, color=isotype_colors[locus], bins=50, alpha=.75)
plt.xlabel("# SHMs in " + locus + "V", fontsize=16)
plt.ylabel("# sequences", fontsize=16)
plt.xticks(fontsize=14)
plt.yticks(fontsize=14)
plt.title('# SHMs in ' + locus + 'V' + ' sequences', fontsize=14)
utils.output_figure(output_fname,
"Distribution of # SHMs in " + locus + "V segments",
log)
def OutputHeatmap(self, output_fname, log):
plt.figure(figsize=(10, 15))
sns.heatmap(np.array(self.abundant_vj_matrix),
xticklabels=self.sorted_js,
yticklabels=self.used_vs,
cmap='jet')
plt.yticks(rotation=0, fontsize=10)
plt.xticks(rotation=90, fontsize=10)
utils.output_figure(
output_fname, "VJ heatmap for the most abundant VJ combinations",
log)
def output_synonymous_shms(synonymous_pos, output_fname, log):
if len(synonymous_pos) < 100:
return
plt.hist(synonymous_pos, color='r', bins=100)
plt.xlabel("Relative position of V SHM in read", fontsize=14)
plt.ylabel("#SHMs", fontsize=14)
plt.xlim(0, .75)
plt.xticks(fontsize=12)
plt.yticks(fontsize=12)
utils.output_figure(
output_fname, "Distribution of synonymous SHM positions in V segment",
log)
def OutputJUsage(self, output_fname, log):
plt.figure(figsize=(10, 8))
perc_list = [
float(self.j_dict[j]) / len(self.vj_df) * 100
for j in self.sorted_js
]
plt.bar(range(len(self.sorted_js)), perc_list)
plt.xticks(range(len(self.sorted_js)),
self.sorted_js,
rotation=90,
fontsize=10)
plt.ylabel('% of sequences', fontsize=14)
utils.output_figure(output_fname, "Usage of J genes", log)
def visualize_region_lengths(labeling_df, region, region_name, output_fname,
log):
region_seq = list(labeling_df[region])
region_len = [len(s) for s in region_seq if len(s) > 1]
f, ax = plt.subplots(figsize=(8, 8))
sns.distplot(region_len, kde=False, rug=False)
plt.xlabel(region_name + ' length', fontsize=16)
plt.ylabel('# ' + region_name + 's', fontsize=16)
plt.xticks(fontsize=14)
plt.yticks(fontsize=14)
plt.xlim(0, 100)
utils.output_figure(output_fname, region_name + " length distribution",
log)
def visualize_largest_group_aa_variability(labeling_df, region, region_name,
output_fname, log):
region_seq = list(labeling_df[region])
max_group = get_region_largest_group(region_seq)
if len(max_group) == 0:
return
group_len = len(max_group[0])
if group_len % 3 != 0:
print "Largest " + region_name + " is not out-of-frame"
return
aa_seqs = [Seq(cdr).translate(to_stop=True) for cdr in max_group]
aa_list = [dict() for i in range(0, group_len / 3)]
for aa_seq in aa_seqs:
for i in range(0, len(aa_seq)):
if aa_seq[i] not in aa_list[i]:
aa_list[i][aa_seq[i]] = 0
aa_list[i][aa_seq[i]] += 1
aa_num = [len(aa) for aa in aa_list]
aa_large_abun = []
aa_large_acid = []
for aa in aa_list:
aa = sorted(aa.items(), key=operator.itemgetter(1), reverse=True)
sum = 0
for i in aa:
sum += i[1]
aa_large_abun.append(float(aa[0][1]) / float(sum) * 100)
aa_large_acid.append(aa[0][0])
aa_set = set()
aa_colors = get_aa_colors()
for aa in aa_large_acid:
aa_set.add(aa)
for aa in aa_set:
x_ = []
abun_ = []
for i in range(0, len(aa_large_abun)):
x_.append(i)
if aa_large_acid[i] == aa:
abun_.append(aa_large_abun[i])
else:
abun_.append(0)
df = pd.DataFrame({'x': x_, 'y': abun_})
sns.barplot(x='x',
y='y',
data=df,
color=aa_colors[amino_acids.index(aa)])
plt.xticks(range(0, len(aa_large_abun)), aa_large_acid, fontsize=14)
plt.yticks(fontsize=14)
plt.xlabel('The most abundant amino acid', fontsize=16)
plt.ylabel('% ' + region_name + 's', fontsize=16)
utils.output_figure(output_fname, region_name + " aa variability", log)
def OutputGeneMutability(gene_mutability_dict, output_fname, gene_type, log):
df_dict = {'Gene': [], 'Mutability': []}
for gene in gene_mutability_dict:
for m in gene_mutability_dict[gene]:
df_dict['Gene'].append(gene)
df_dict['Mutability'].append(m)
plt.figure(figsize=(10, 8))
sns.boxplot(x='Gene', y='Mutability', data=df_dict)
max_mutability = max(0.55, max(df_dict['Mutability']))
plt.ylim(-0.05, max_mutability)
plt.xticks(rotation=90)
plt.ylabel('Mutability')
utils.output_figure(output_fname, "Mutability of " + gene_type + ' genes',
log)
def OutputGeneSHMPlot(gene_shms, gene_name, gene_length, num_aligned_seqs,
output_fname, log):
nucl_dict = {
'A': [0] * gene_length,
'C': [0] * gene_length,
'G': [0] * gene_length,
'T': [0] * gene_length
}
num_shms = 0
for shm in gene_shms:
if not shm.is_substitution() or not nucl_is_valid(shm.read_nucl):
continue
nucl_dict[shm.read_nucl][shm.gene_pos] += 1
num_shms += 1
for nucl in nucl_dict:
for i in range(len(nucl_dict[nucl])):
nucl_dict[nucl][i] = nucl_dict[nucl][i]
x = range(gene_length)
plt.figure()
plt.bar(x, [
float(sum(y)) / num_aligned_seqs for y in zip(
nucl_dict['A'], nucl_dict['C'], nucl_dict['G'], nucl_dict['T'])
],
color='blue',
label='A')
plt.bar(x, [
float(sum(y)) / num_aligned_seqs
for y in zip(nucl_dict['C'], nucl_dict['G'], nucl_dict['T'])
],
color='green',
label='C')
plt.bar(x, [
float(sum(y)) / num_aligned_seqs
for y in zip(nucl_dict['G'], nucl_dict['T'])
],
color='red',
label='G')
plt.bar(x, [float(m) / num_aligned_seqs for m in nucl_dict['T']],
color='orange',
label='T')
plt.legend(loc='upper center', ncol=4)
plt.ylim(0, 1.1)
plt.xlabel('Position in V gene', fontsize=14)
plt.ylabel('Fraction of sequences', fontsize=14)
plt.title(
str(num_aligned_seqs) + ' sequences were aligned to ' + gene_name)
utils.output_figure(output_fname, "SHM position in " + gene_name, log)
return nucl_dict
def output_shms_pos(all_shms_pos, colors, output_prefix, log):
for isotype in all_shms_pos:
if len(all_shms_pos[isotype]) < 10:
continue
plt.hist(all_shms_pos[isotype],
bins=100,
color=colors[isotype],
alpha=.75)
plt.xlabel("#SHM in " + isotype + "V gene segment", fontsize=16)
plt.ylabel("# sequences", fontsize=16)
plt.xticks(fontsize=14)
plt.yticks(fontsize=14)
plt.xlim(0, .75)
output_fname = output_prefix + "_" + isotype + "V_pos"
utils.output_figure(
output_fname, "Distribution of SHM relative positions in " +
isotype + "V segments", log)
def output_num_shms(num_all_shms, colors, output_prefix, log):
pos = []
labels = []
cols = []
for isotype in num_all_shms:
if len(num_all_shms[isotype]) > 0:
pos.append(num_all_shms[isotype])
labels.append(str(isotype))
cols.append(colors[isotype])
plt.hist(pos, bins=50, color=cols, alpha=.75, label=labels)
plt.legend(loc='upper center', ncol=len(pos), fontsize=16)
plt.xlabel("# of SHM in V gene segment", fontsize=16)
plt.ylabel("# SHMs", fontsize=16)
plt.xticks(fontsize=14)
plt.yticks(fontsize=14)
plt.xlim(0, 150)
output_fname = output_prefix + "_shms_number"
utils.output_figure(output_fname, "Distribution of # SHMs in V segments",
log)
def visualize_vj_heatmap(labeling_df, output_pdf, log):
v_hits = list(labeling_df['V_hit'])
j_hits = list(labeling_df['J_hit'])
if len(v_hits) == 0 or len(j_hits) == 0:
log.info("VJ data-frame contains 0 records. VJ usage visualization will be skipped")
return
vj_matrix = VJMatrix(v_hits, j_hits)
log.info(str(len(vj_matrix.vj_dict)) + " VJ pairs were extracted. Pairs are presented by " +
str(len(vj_matrix.v_set)) + " V genes & " + str(len(vj_matrix.j_set)) + " J genes")
table, v, j = vj_matrix.CreateTable(100)
mplt.rcParams.update({'font.size': 20})
#plt.figure(figsize=(15, 15))
f, ax = plt.subplots(figsize=(10, 15))
sns.heatmap(table, cmap = plt.cm.jet, xticklabels = v, yticklabels = j, ax = ax)
ax.tick_params(labelsize = 16)
x = [i + 0.0 for i in range(0, len(v))]
y = [i + .5 for i in range(0, len(j))]
plt.xticks(x, v, rotation=60, fontsize=14)
plt.yticks(y, j, rotation='horizontal', fontsize=14)
utils.output_figure(output_pdf, "VJ heatmap for the most abundant VJ combinations", log)
def visualize_aa_substitution_matrix(shms_df, output_fname, log):
dict_aa = dict()
num_shms = 0
for it in shms_df:
read_shms = shms_df[it]
prev_pos = -1
for shm in read_shms:
if prev_pos / 3 != shm.read_pos / 3:
if aa_is_valid(shm.gene_aa) and aa_is_valid(shm.read_aa):
aa_pair = shm.gene_aa + shm.read_aa
if not aa_pair in dict_aa:
dict_aa[aa_pair] = 0
dict_aa[aa_pair] += 1
num_shms += 1
prev_pos = shm.read_pos
aa_list = get_aa_list()
aa_freq = []
for i in range(0, len(aa_list)):
aa_freq.append([0] * len(aa_list))
for aa_pair in dict_aa:
aa_freq[aa_list.index(aa_pair[1])][aa_list.index(
aa_pair[0])] = float(dict_aa[aa_pair]) / float(num_shms)
fig, ax = plt.subplots()
sns.heatmap(aa_freq,
cmap=plt.cm.jet,
xticklabels=aa_list,
yticklabels=aa_list,
square=True,
ax=ax)
ax.tick_params(labelsize=14)
plt.xticks(fontsize=12)
plt.yticks(fontsize=12, rotation='horizontal')
#for tick, color in zip(ax.get_xticklabels(), get_aa_ticks_colors(aa_list)):
# tick.set_color(color)
#for tick, color in zip(ax.get_yticklabels(), get_aa_ticks_colors(aa_list)):
# tick.set_color(color)
plt.xlabel("To", fontsize=14)
plt.ylabel("From", fontsize=14, rotation='horizontal')
utils.output_figure(output_fname, "Amino acid substitution heatmap", log)
return aa_freq
def OutputFractionOfSynonymousSHMs(shm_df, output_fname, log):
v_shm_fractions = []
j_shm_fractions = []
for it in shm_df:
read_shms = shm_df[it]
num_synonymous = 0
for shm in read_shms:
if shm.synonymous:
num_synonymous += 1
fraction = 0
if len(read_shms) != 0:
fraction = float(num_synonymous) / len(read_shms)
if it.is_variable():
v_shm_fractions.append(fraction)
else:
j_shm_fractions.append(fraction)
plt.hist([v_shm_fractions, j_shm_fractions], label=['V gene', 'J gene'])
plt.xlabel('Fraction of synonymous SHMs', fontsize=14)
plt.ylabel('# sequences', fontsize=14)
plt.legend(loc='upper right', fontsize=14)
utils.output_figure(output_fname,
"Fractions of synonymous SHMs in V and J genes", log)
