def poly_aaa_vs_spidex(variants_by_gene):
"""Analysis of poly A track changing mutations using data from SPIDEX."""
aaa_variants_list = all_poly_a_variants(variants_by_gene)
raw_report = spidex_from_list(aaa_variants_list)
print('Unique points', len(raw_report))
print('Plotting')
variants_groups = divide_variants_by_poly_aaa(raw_report)
plot_aaa_vs_spidex(variants_groups)
print('ks test')
spidex_aaa_ks_test(variants_groups, already_divided=True)
def get_data_from_ensembl_api(variants):
import requests
api_report = []
for variant in all_poly_a_variants(variants,
merge_variants_with_multiple_id=False):
server = 'http://rest.ensembl.org'
# server = 'http://grch37.rest.ensembl.org/' GRCH 37 has no eqtls implemented
ext = '/eqtl/variant_name/homo_sapiens/%s?statistic=p-value;content-type=application/json' % variant.snp_id
try:
r = requests.get(server + ext,
headers={'content-type': 'application/json'})
if not r.ok:
r.raise_for_status()
sys.exit()
decoded = r.json()
if 'error' not in decoded:
print('Got data for %s' % variant.snp_id)
# print(repr(decoded))
for datum in decoded:
for transcript in variant.affected_transcripts:
for alt, aaa_data in transcript.poly_aaa.items():
report_chunk = (variant.snp_id, datum['tissue'],
datum['value'], datum['gene'],
aaa_data.increased,
aaa_data.decreased,
aaa_data.change, variant.chr_name,
variant.chr_start, variant.ref,
alt, transcript.strand,
transcript.ensembl_id,
transcript.cds_start,
transcript.cds_end)
api_report += [report_chunk]
except Exception as e:
print(e)
return api_report
def poly_aaa_vs_expression(variants_by_gene):
bdb = ExpressionDatabase(GTEX_DATABASE)
def is_length_difference_big(l1, l2):
"""Is the first list much longer than the second?"""
len1 = len(l1)
len2 = len(l2)
assert len1 > len2
if len2 == 0 or len1 // len2 > 10:
return True
gtex_report = []
gtex_report_with_tissue = []
aaa_variants_list = list(all_poly_a_variants(variants_by_gene))
print('Analysing %s poly_a related variants (out of %s total).' %
(len(aaa_variants_list), len(variants_by_gene)))
for variant in aaa_variants_list:
for transcript in variant.affected_transcripts:
if not transcript.poly_aaa:
continue
expression_data_by_alt = bdb.get_by_mutation(variant, transcript)
transcript.expression = {}
for alt, aaa_data in transcript.poly_aaa.items():
expression_data = expression_data_by_alt.get(alt, None)
if not expression_data:
continue
else:
print('Expression data for', variant.snp_id, 'found:',
expression_data)
expression_up = []
expression_down = []
data = transcript.poly_aaa[alt]
for tissue_name, slope, gene in expression_data:
gtex_report_with_tissue.append(
(variant.snp_id, tissue_name, slope, gene,
data.increased, data.decreased, data.change,
variant.chr_name, variant.chr_start, variant.ref, alt,
transcript.strand, transcript.ensembl_id,
transcript.cds_start, transcript.cds_end))
slope = float(slope)
if slope > 0:
expression_up += [tissue_name]
elif slope < 0:
expression_down += [tissue_name]
# is this rather up?
if len(expression_up) > len(expression_down):
# is this certainly up?
if is_length_difference_big(expression_up,
expression_down):
expression_trend = 'up'
else:
expression_trend = 'rather_up'
# is this rather down?
elif len(expression_down) > len(expression_up):
# is this certainly down?
if is_length_difference_big(expression_down,
expression_up):
expression_trend = 'down'
else:
expression_trend = 'rather_down'
# is unchanged?
else:
expression_trend = 'constant'
expression_up_in_x_cases = len(expression_up)
expression_down_in_x_cases = len(expression_down)
transcript.expression[alt] = expression_trend
report_chunk = (variant.snp_id, expression_up_in_x_cases,
expression_down_in_x_cases, expression_trend,
data.increased, data.decreased, data.change,
variant.chr_name, variant.chr_start,
variant.ref, alt, transcript.strand,
transcript.ensembl_id, transcript.cds_start,
transcript.cds_end)
gtex_report += [report_chunk]
"""
gtex_report += [(
sum('up' in v.expression.values() for v in poly_a_related_variants),
sum('down' in v.expression.values() for v in poly_a_related_variants),
sum(
sum('up' == expr for expr in v.expression.values())
for v in poly_a_related_variants
),
sum(
sum('down' == expr for expr in v.expression.values())
for v in poly_a_related_variants
),
sum(data.increased for v in poly_a_related_variants for data in v.poly_aaa.values()),
sum(data.decreased for v in poly_a_related_variants for data in v.poly_aaa.values())
)]
"""
report('expression table for variants (based on data from gtex)',
['\t'.join(map(str, line)) for line in gtex_report], [
'variant', 'expression+', 'expression-', 'trend', 'aaa+',
'aaa-', 'aaa_change', 'chrom', 'pos', 'ref', 'alt', 'strand',
'transcript', 'cds_start', 'cds_end'
])
report(
'expression table for variants with tissues (based on data from gtex)',
['\t'.join(map(str, line)) for line in gtex_report_with_tissue], [
'variant', 'tissue', 'slope', 'gene', 'aaa+', 'aaa-', 'aaa_change',
'chrom', 'pos', 'ref', 'alt', 'strand', 'transcript', 'cds_start',
'cds_end'
])
summarize_tissue_eqtl_aaa_correlation(gtex_report_with_tissue)
#report(
# 'Expression table for genes (based on data from GTEx)',
# ['\t'.join(map(str, line)) for line in gtex_report_by_genes],
# # note: alleles is not the same as variants
# [
# 'gene', 'alleles with expression+', 'alleles with expression-',
# 'variants with expression+', 'variants with expression-', '#aaa+', '#aaa-'
# ]
#)
print('Done')
def spidex_aaa_ks_test(variants_groups, already_divided=False):
if not already_divided:
aaa_variants_list = all_poly_a_variants(variants_groups)
raw_report = spidex_from_list(aaa_variants_list)
variants_groups = divide_variants_by_poly_aaa(raw_report)
groups_zscores = {
name: [point['dpsi_zscore'] for point in group]
for name, group in variants_groups.items()
}
for group_1, group_2 in combinations(groups_zscores, 2):
print('%s vs %s:' % (group_1, group_2))
z_scores_1 = groups_zscores[group_1]
z_scores_2 = groups_zscores[group_2]
ks_result = ks_2samp(z_scores_1, z_scores_2)
print(ks_result)
groups_new_aaa_lengths = defaultdict(list)
for name, group in variants_groups.items():
for point in group:
new_aaa_length = point['new_aaa_length']
groups_new_aaa_lengths[new_aaa_length].append(point['dpsi_zscore'])
group = None
name = None
ks_results = {}
for new_aaa_length in sorted(groups_new_aaa_lengths):
print(
'All mutations causing poly_aaa to be <= %s vs all mutations causing poly_aaa to be > %s:'
% (new_aaa_length, new_aaa_length))
z_scores_1 = [
zscore for name, group in groups_new_aaa_lengths.items()
for zscore in group if name <= new_aaa_length
]
z_scores_2 = [
zscore for name, group in groups_new_aaa_lengths.items()
for zscore in group if name > new_aaa_length
]
if not z_scores_2:
print('No mutations causing poly_aaa to be > %s' % new_aaa_length)
continue
ks_result = ks_2samp(z_scores_1, z_scores_2)
print(new_aaa_length, ks_result)
ks_results[new_aaa_length] = -np.log(ks_result.pvalue)
lengths = list(ks_results.keys())
plt.hist(lengths,
weights=list(ks_results.values()),
bins=list(ks_results.keys()),
rwidth=0.9)
plt.xticks(lengths)
plt.xlabel('Length of poly(A) track: $x$')
plt.ylabel(r'$-\log($P-Value$)$')
plt.title(
'Ks-test for groups: '
'mutations effecting in poly(A) length $\leq$ $x$ vs mutations effecting in poly(A) length > $x$'
)
plt.grid(True)
save_plot(plt)
def summarize_poly_aaa_variants(variants):
columns = [
'snp_id', 'gene', 'poly_aaa_increase', 'poly_aaa_decrease',
'poly_aaa_change', 'chr', 'start', 'end', 'ref', 'alt', 'transcript',
'cds_start', 'cds_end'
]
Record = recordclass('RecordPolyA', columns)
aaa_records = []
aaa_variants = set()
up_variants = {}
down_variants = {}
all_variants_ids = []
variants_sources = Counter()
transcripts = set()
new_poly_a = 0
in_poly_a = 0
for variant in all_poly_a_variants(variants, preserve_sources=True):
all_variants_ids.extend(variant.snp_id.split(','))
new = False
in_a = False
for transcript in variant.affected_transcripts:
if not transcript.poly_aaa:
continue
for alt, aaa_data in transcript.poly_aaa.items():
record = Record(
variant.snp_id,
None, #variant.ensembl_gene_stable_id # TODO
aaa_data.increased,
aaa_data.decreased,
aaa_data.change,
variant.chr_name,
variant.chr_start,
variant.chr_end,
variant.ref,
alt,
transcript.ensembl_id,
transcript.cds_start,
transcript.cds_end)
if not aaa_data.has and aaa_data.will_have:
new = True
if aaa_data.has:
in_a = True
if aaa_data.increased:
up_variants[variant] = True
if aaa_data.decreased:
down_variants[variant] = True
transcripts.add(transcript.ensembl_id)
aaa_records.append(record)
aaa_variants.add(variant)
if new:
new_poly_a += 1
if in_a:
in_poly_a += 1
for source in set(variant.source.split(',')):
variants_sources[source] += 1
report('poly aaa increase and decrease by variants', aaa_records, columns)
report('poly aaa sources', variants_sources.items(), ['source', 'count'])
report('all ids', all_variants_ids)
print('Variants creating new poly(A) tracks: %s' % new_poly_a)
print('Variants in existing poly(A) tracks: %s' % in_poly_a)
print('Affected transcripts: %s' % len(transcripts))
print('Down variants: %s' % len(down_variants))
print('Up variants: %s' % len(up_variants))
print('Unique variants: %s' % len(aaa_variants))
print('Variants identifiers: %s' %
sum(v.snp_id.count(',') + 1 for v in aaa_variants))
print(variants_sources)
def poly_aaa_consequences(variants):
mutations_in_cds_hgvs_format = defaultdict(list)
indels = Counter()
all = Counter()
for variant in all_poly_a_variants(variants, preserve_sources=True):
for transcript in variant.affected_transcripts:
if not transcript.poly_aaa:
continue
for alt, aaa_data in transcript.poly_aaa.items():
if aaa_data.increased:
category = 'increased'
elif aaa_data.decreased:
category = 'decreased'
else:
category = 'constant'
hgvs = transcript.as_hgvs(variant.ref, alt)
if 'del' in hgvs or 'ins' in hgvs:
indels[category] += 1
all[category] += 1
mutations_in_cds_hgvs_format[category].append(hgvs)
mutations_in_cds_hgvs_format['all'].append(hgvs)
print('Indels enrichment:')
for category in indels:
print(category, indels[category] / all[category] * 100, '%')
for category, muts in mutations_in_cds_hgvs_format.items():
report(
'Mutations which result in ' + category + ' in cds hgvs formats',
muts)
consequences = defaultdict(Counter)
skipped = Counter()
for category, muts in mutations_in_cds_hgvs_format.items():
filename = report(
'Mutations which result in ' + category + ' in cds hgvs formats',
muts)
vep_filename = vep(filename)
for line in open(vep_filename):
if line.startswith('#'):
continue
line = line.split('\t')
tested_transcript = line[0].split(':')[0]
vep_transcript = line[4]
if line[5] != 'Transcript':
skipped['Not a transcript feature'] += 1
continue
if tested_transcript != vep_transcript:
skipped['Different transcript'] += 1
continue
variant_consequences = line[6].split(',')
for consequence in variant_consequences:
consequences[category][consequence] += 1
print(skipped)
print('Raw consequences')
print(consequences)
graph = load_sequence_ontology()
expanded_consequences = propagate_consequences(graph, consequences)
for category, counts in expanded_consequences.items():
consequences_to_include = ['coding_sequence_variant']
consequences_to_include.extend(counts.keys())
g = graph.subgraph([
node for node, data in graph.nodes(data=True)
if data['name'] in consequences_to_include
])
g = g.reverse()
max_count = max(counts.values())
for node, data in g.nodes_iter(data=True):
name = data['name']
count = counts[name]
color = (255 - int(log((count / max_count) + 1) * 255), 255, 255)
g.node[node]['style'] = 'filled'
g.node[node]['shape'] = 'box'
color = '#%02x%02x%02x' % color
g.node[node]['fillcolor'] = color
if name not in consequences[category]:
g.node[node]['style'] = 'dashed,filled'
g = nx.relabel_nodes(
g, {
node: data['name'].replace('variant', 'v.') +
': %s' % counts.get(data['name'])
for node, data in g.nodes(data=True)
})
a = nx_agraph.to_agraph(g)
a.layout(
'dot',
args=
'-Nfontsize=14 -Nwidth=".2" -Nheight=".2" -Nmargin=.1 -Gfontsize=8 -Earrowsize=.5'
)
a.draw('reports/poly_a_consequences_dag_' + category + '.svg')
selected_consequences_groups = {
'General coding':
['synonymous_variant', 'frameshift_variant', 'inframe_variant'],
'Inframe': [
'inframe_deletion', 'inframe_insertion', 'missense_variant',
'stop_gained', 'stop_lost'
]
}
for group, selected_consequences in selected_consequences_groups.items():
for category, counts in expanded_consequences.items():
data = {
consequence: counts[consequence]
for consequence in selected_consequences
}
data = OrderedDict(sorted(data.items(), key=itemgetter(1)))
# Create a pie chart
wedges = plt.pie(
list(data.values()),
labels=list(data.keys()),
shadow=False,
colors=plt.cm.tab20(
numpy.linspace(1, 0, len(selected_consequences))),
startangle=0,
autopct='%1.1f%%',
)
for pie_wedge in wedges[0]:
pie_wedge.set_edgecolor('black')
# View the plot drop above
plt.axis('equal')
plt.title(group + ' consequences for variants causing ' +
category + ' in poly(A) length')
plt.tight_layout()
save_plot(plt, hide_title=True)