def main():
global no_neighbor_count
with safe_open(outpath, exist_ok='exit') as outfile:
# write header
#outfile.write('\t'.join(['head_id', 'gene_id', 'flag', 'distance'])+'\n')
relations = pd.DataFrame()
for head_group_name, head_group in tqdm(head_groups, desc='Contigs'):
if head_group_name not in gene_groups.groups:
prinf(
'Não há nenhum gene no cromossomo. As heads abaixo não possuem NG.'
)
prinf(head_group)
no_neighbor_count += head_group.shape[0]
continue
gene_group = gene_groups.get_group(head_group_name)
chunks = np.array_split(head_group, n_cpu)
with mp.Pool() as pool:
pool_results = pool.starmap(parse_chunk,
((c, gene_group, cn)
for cn, c in enumerate(chunks)))
for chunk_relations in pool_results:
relations = relations.append(chunk_relations)
# print('\nCHUN', chunk_relations, '\nREL', relations)
relations.columns = ['head_id', 'gene_id', 'flag', 'distance']
relations.to_csv(outfile, sep='\t', index=False)
log(f'\nConcluído. Relações salvas em {str(outpath)}.')
return relations, no_neighbor_count
def parse_head_row(head_row, gene_group, outfile):
parse_head_row.last_args = head_row, gene_group, outfile
for _, gene_row in gene_group.iterrows():
if overlaps((gene_row.start, gene_row.end),
(head_row.start, head_row.end)):
flag = 'olap'
chosen_gene_id = gene_row.id
distance = 0
break
# if none overlaps
else:
# b f de forward ou backfill
distance = 9e99 # create Inf() class for that?
for meth in ['b', 'f']:
fill_back = meth == 'b'
gene_pos = ['end', 'start'][fill_back]
head_pos = ['end', 'start'][not fill_back]
if gene_group[gene_pos].duplicated().sum():
raise pd.core.indexes.base.InvalidIndexError(
'You appear to have duplicated gene positions. '
'Indexes won\'t work properly.')
indexed_group = gene_group.sort_values(gene_pos).set_index(
gene_pos)
try:
near_gene_index = indexed_group.index.get_loc(
head_row[head_pos], meth + 'fill')
gene_row = gene_group.iloc[near_gene_index]
new_distance = abs(gene_row[gene_pos] - head_row[head_pos])
if new_distance < distance:
chosen_gene_id = gene_row.id
flag = ["gh", "hg"][fill_back]
distance = new_distance
except KeyError as e:
prinf(
f'\nNão há gene {meth.upper()} de {head_row.id}. (B=atrás, F=à frente)\n'
)
outfile.write('\t'.join([head_row.id, chosen_gene_id, flag,
str(distance)]) + '\n')
def main():
filtered_outfile = safe_open(filtered_outpath, exist_ok=False)
discarded_outfile = safe_open(discarded_outpath, exist_ok=False)
n_cpu = mp.cpu_count()
#================== LER E FILTRAR ALINHAMENTOS ==================#
print('Lendo resultados do Blast...', end=' ')
perere3_vs_genoma = pd.read_table(perere3_inpath,
header=None,
names=BL_COLUMNS)
sr3_vs_genoma = pd.read_table(sr3_inpath, header=None, names=BL_COLUMNS)
print('Resultados lidos.')
# Sort positions
# for data in (perere3_vs_genoma, sr3_vs_genoma):
# data.sort_values('sstart', inplace=True)
# data.reset_index(drop=True, inplace=True)
print('Buscando alinhamentos em que o SR3 é melhor...')
discarded = pd.DataFrame()
filtered_perere3_vs_genoma = perere3_vs_genoma.copy()
p_groups = perere3_vs_genoma.groupby('saccver')
s_groups = sr3_vs_genoma.groupby(
'saccver') ## agrupar muda index???????????
print(
'Iterando para cada scaffold no genoma e para cada perere3 no scaffold.'
)
for p_group_name in tqdm(p_groups.groups, desc='Scaffolds'):
s_group = s_groups.get_group(p_group_name)
p_group = p_groups.get_group(p_group_name)
prinf('Combinando DataFrames...', end='\r')
product = cartesian_product(
p_group[['sstart', 'send', 'bitscore']].reset_index(),
s_group[['sstart', 'send', 'bitscore']].reset_index())
# discard when perere3 aligns better
prinf('Filtrando por bitscore do SR3...', end='\r')
product = product.loc[product.bitscore_x < product.bitscore_y]
if product.empty:
continue
prinf('Subdividindo produto... ', end='\r')
product_chunks = np.array_split(product, n_cpu)
prinf('Procurando sobreposições...', end='\r')
with mp.Pool() as pool:
chunks_discarded = pool.starmap(parse_product,
enumerate(product_chunks))
group_discarded = pd.concat(chunks_discarded)
discarded = discarded.append(group_discarded)
# print(discarded[~discarded['index'].isin(filtered_perere3_vs_genoma.index)])
# print(filtered_perere3_vs_genoma.loc[discarded['index'].unique()])
print(
f"Escrevendo posições das linhas removidas de '{str(perere3_inpath)}' em '{str(discarded_outpath)}'...",
end=' ')
discarded.columns = pd.MultiIndex.from_product([('perere3', 'sr3'),
('index', 'sstart',
'ssend', 'bitscore')])
discarded.to_csv(discarded_outfile, sep='\t', index=False)
print('Arquivo escrito.')
print('Filtrando...', end=' ')
filtered_perere3_vs_genoma.drop(discarded[('perere3', 'index')],
inplace=True)
print(f'\nFiltragem concluída. {len(discarded)} alinhamentos removidos.')
print(
f"Escrevendo alinhamentos filtrados do perere3 em '{str(filtered_outpath)}'...",
end=' ')
filtered_perere3_vs_genoma.to_csv(filtered_outfile, sep='\t', index=False)
print('Arquivo escrito.')
return filtered_perere3_vs_genoma, discarded
print('Concluído. Calculando correlações...')
gridx, gridy = 5, 8
plt.rcParams['font.size'] = 4
plt.rcParams['figure.figsize'] = (16, 9)
i = 0
for _, relation_row in relations.iterrows():
hid = relation_row.head_id
gid = relation_row.gene_id
if hid not in counts:
prinf(
f'WARNING: {hid} não presente nas contagens, só nas relações head-gene. Talvez a contagem deva ser refeita.'
)
continue
head_col = counts[hid]
gene_col = counts[gid]
n_non_zero = (head_col.astype(bool) & gene_col.astype(bool)).sum()
# if number of significative points (head and gene counts != 0) is less than 4
if n_non_zero < 4:
prinf('\nCorrelação descartada:\n', counts[[gid, hid]])
continue
prinf(
'\nCorrelação plotada:\n',
pd.concat([
counts[[gid, hid]],
# genes.loc[genes.duplicated('end', 'last'), 'end'] += 1
gene_groups = genes.groupby(COLS_TO_GROUP)
if __name__ == '__main__':
# write header
outfile.write('\t'.join(['head_id', 'gene_id', 'flag', 'distance']) + '\n')
print('Iterate for each contig.')
for head_group_name, head_group in tqdm(head_groups):
try:
gene_group = gene_groups.get_group(head_group_name)
except KeyError:
prinf(
'Não há nenhum gene no cromossomo. As heads abaixo são "desgenadas".'
)
prinf(head_group)
continue
# parse head_row for each head_row
head_group_chunks = pd.np.array_split(head_group, mp.cpu_count())
def parse_chunk(df):
global n_cpu
tqdm.pandas(position=parse_chunk.bar_pos + 1)
parse_chunk.bar_pos += 1
print(parse_chunk.bar_pos, '****', n_cpu)
parse_chunk.bar_pos %= n_cpu
return df.progress_apply(parse_head_row,
axis=1,
genes['id'] = parse_gff_attributes(genes.attributes).index
head_groups = heads.groupby(COLS_TO_GROUP)
gene_groups = genes.groupby(COLS_TO_GROUP)
if __name__ == '__main__':
# write header
outfile.write('\t'.join(['head_id', 'gene_id', 'flag', 'distance']) + '\n')
print('Iterate for each contig and for each head in contig.')
for head_group_name, head_group in tqdm(head_groups):
try:
gene_group = gene_groups.get_group(head_group_name)
except KeyError:
prinf(
'Não há nenhum gene no cromossomo. As heads abaixo são "desgenadas".'
)
prinf(head_group)
continue
for _, head_row in tqdm(list(head_group.iterrows())):
for _, gene_row in gene_group.iterrows():
if overlaps((gene_row.start, gene_row.end),
(head_row.start, head_row.end)):
flag = 'olap'
chosen_gene_id = gene_row.id
distance = 0
break
# if none overlaps
def main():
heads = pd.read_table(head_annotations_path,
names=GFF3_COLUMNS,
usecols=GFF_COLS_SUBSET)
genes = pd.read_table(gene_annotations_path,
names=GFF3_COLUMNS,
usecols=GFF_COLS_SUBSET)
heads['id'] = parse_gff_attributes(heads.attributes).index
genes['id'] = parse_gff_attributes(genes.attributes).index
head_groups = heads.groupby(COLS_TO_GROUP)
gene_groups = genes.groupby(COLS_TO_GROUP)
outfile = safe_open(outpath, exist_ok=False)
# write header
outfile.write('\t'.join(['head_id', 'gene_id', 'flag', 'distance']) + '\n')
print('Iterate for each contig and for each head in contig.')
for head_group_name, head_group in tqdm(head_groups):
try:
gene_group = gene_groups.get_group(head_group_name)
except KeyError:
prinf(
'Não há nenhum gene no cromossomo. As heads abaixo são "desgenadas".'
)
prinf(head_group)
continue
for _, head_row in tqdm(list(head_group.iterrows())):
for _, gene_row in gene_group.iterrows():
if overlaps((gene_row.start, gene_row.end),
(head_row.start, head_row.end)):
flag = 'olap'
chosen_gene_id = gene_row.id
distance = 0
break
# if none overlaps
else:
# b f de forward ou backfill
distance = 9e99 # create Inf() class for that?
for meth in ['b', 'f']:
fill_back = meth == 'b'
gene_pos = ['end', 'start'][fill_back]
head_pos = ['end', 'start'][not fill_back]
if gene_group[gene_pos].duplicated().sum():
raise pd.core.indexes.base.InvalidIndexError(
'You appear to have duplicated gene positions.'
'Indexes won\'t work properly.')
indexed_group = gene_group.sort_values(gene_pos).set_index(
gene_pos)
try:
near_gene_index = indexed_group.index.get_loc(
head_row[head_pos], meth + 'fill')
gene_row = gene_group.iloc[near_gene_index]
new_distance = abs(gene_row[gene_pos] -
head_row[head_pos])
if new_distance < distance:
chosen_gene_id = gene_row.id
flag = ["gh", "hg"][fill_back]
distance = new_distance
except KeyError:
prinf(f'Não há gene à {flag} de {head_row.id}.')
outfile.write(
'\t'.join([head_row.id, chosen_gene_id, flag,
str(distance)]) + '\n')
print(f'\nConcluído. Relações salvas em {str(outpath)}.')
outfile.close()
def main():
u.log(f'{__file__}: Generating heads of {HEAD_LEN} bp.')
truncated_count = 0
#======================== LEITURA ========================#
heads_annotations_file = u.safe_open(heads_annotations_path,
exist_ok=False)
heads_outfile = u.safe_open(heads_outpath, exist_ok=False)
motherlength_outfile = u.safe_open(motherlength_path, exist_ok=False)
print('Lendo alinhamentos filtrados do Perere3...', end=' ')
filtered_perere3_vs_genoma = read_table(inpath)
print(f"'{inpath}' lido.")
print('Lendo genoma de S. mansoni...', end=' ')
genomedict = to_dict(parse(str(u.genome_path), 'fasta'))
print('Dicionário criado.')
#======================== GET HEADS ========================#
print('Searching for Perere-3 copies in S. mansoni\'s genome...')
with (u.pardir / 'seqs/perere3.fa').open() as per_file:
perere_len = len(''.join([l.strip()
for l in per_file.readlines()][1:]))
heads = []
for index, row in filtered_perere3_vs_genoma.iterrows():
# Discard copies without 3' end.
if abs(row['qend'] - perere_len) < MAX_DISTANCE_FROM_END:
genome_piece = genomedict[row['saccver']].seq
plus_sense = row['sstart'] < row['send']
if plus_sense:
head_slice = slice(row['send'], row['send'] + GTAA_WINDOW_LEN)
proto_head = genome_piece[head_slice]
if u.verbose:
prefix = genome_piece[head_slice.start -
PREFIX_LEN:head_slice.start]
else:
head_slice = slice(row['send'] - GTAA_WINDOW_LEN - 1,
row['send'] - 1)
proto_head = genome_piece[head_slice].reverse_complement()
if u.verbose:
prefix = genome_piece[head_slice.stop:head_slice.stop +
PREFIX_LEN].reverse_complement()
###### Algumas dão xabu (???)
if head_slice.start < 0 or head_slice.stop < 0:
u.prinf(f'Head descartada com posições:', head_slice)
continue
skip_gtaa = find_gtaa_break(proto_head)
head = proto_head[skip_gtaa:skip_gtaa + HEAD_LEN]
#======================== ANOTAR HEAD NO GFF ========================#
if plus_sense:
start_pos = row['send'] + 1 + skip_gtaa
end_pos = start_pos + HEAD_LEN
else:
end_pos = row['send'] - 1 - skip_gtaa
start_pos = end_pos - HEAD_LEN
heads_annotations_file.write('\t'.join([
row['saccver'], 'WormBase_imported', 'gene',
str(start_pos),
str(end_pos), '.', ['-', '+'][plus_sense], '.',
f'gene_id=head{index};motherlength={row["length"]};length={HEAD_LEN}'
]) + '\n')
motherlength_outfile.write(f"head{index}\t{row['length']}\n")
#======================== ESCREVER HEAD EM FASTA ========================#
heads_outfile.write(f'>head{index}\n' + str(head) + '\n')
#========================================================================#
if u.verbose:
print(
['-', '+'][plus_sense], prefix, proto_head[:skip_gtaa] +
' | ' + head[:30 - skip_gtaa] + '...',
f" {len(head)}bp\t{row['pident']:.2f}%\t{row['evalue']:.2e}\t{row['bitscore']:5}\t{row['saccver']}\t{head_slice.start}-{head_slice.stop}"
)
else:
truncated_count += 1
u.log(
f'\n{filtered_perere3_vs_genoma.shape[0] - truncated_count} heads written:',
heads_outpath,
heads_annotations_path,
sep='\n\t')
u.log(f'{filtered_perere3_vs_genoma.shape[0]} alignments considered.')
u.log(truncated_count, 'heads discarded as truncated.')
heads_annotations_file.close()
heads_outfile.close()
motherlength_outfile.close()
#======================== PLOTAR HISTOGRAMAS ========================#
if u.plot_flag:
heads_df = DataFrame.from_dict(dict(enumerate(zip(*heads))))
for j in range(8):
plt.figure(figsize=(16, 9))
for i in range(12):
plt.subplot(3, 4, i + 1)
heads_df[i + j * 12].value_counts().plot(kind='bar')
plt.show()
