def parse_stat_fname(self, stat_fname, filter_type):
sys.stderr.write("Parsing stat_fname: %s ...\n"%os.path.basename(stat_fname))
reader = csv.reader(open(stat_fname), delimiter='\t')
prediction_pair2instance = {}
#following is temp default values
p_gene_id = -2 #-2 means no database submission for p_gene_id
avg_p_value = -1
no_of_clusters = 1
p_value_cut_off = -1
connectivity_cut_off = -1
cluster_size_cut_off = -1
unknown_cut_off = -1
vertex_gradient = -1
counter = 0
real_counter = 0
for row in reader:
cluster_id, gene_no, go_no, go_no_depth, recurrence, gradient_score, edge_gradient, \
is_correct, is_correct_L1, is_correct_lca, lca_list_string = row
cluster_id = int(cluster_id)
gene_no = int(gene_no)
go_no = int(go_no)
go_no_depth = int(go_no_depth)
recurrence = float(recurrence)
edge_gradient = float(edge_gradient)
is_correct = int(is_correct)
is_correct_L1 = int(is_correct_L1)
is_correct_lca = int(is_correct_lca)
cluster_array = '{%s}'%cluster_id
new_row = [p_gene_id, gene_no, go_no, is_correct, is_correct_L1, \
is_correct_lca, avg_p_value, no_of_clusters, cluster_array, p_value_cut_off, recurrence, \
connectivity_cut_off, cluster_size_cut_off, unknown_cut_off, go_no_depth, cluster_id, \
lca_list_string, vertex_gradient, edge_gradient]
p_attr_instance = prediction_attributes(new_row, type=2)
prediction_pair = (gene_no, go_no)
if prediction_pair not in prediction_pair2instance:
prediction_pair2instance[prediction_pair] = p_attr_instance
real_counter += 1
else: #remove redundancy
if filter_type==1:
new_cmp_value = p_attr_instance.recurrence_cut_off
old_cmp_value = prediction_pair2instance[prediction_pair].recurrence_cut_off
elif filter_type==2:
new_cmp_value = p_attr_instance.edge_gradient
old_cmp_value = prediction_pair2instance[prediction_pair].edge_gradient
elif filter_type==3:
new_cmp_value = p_attr_instance.recurrence_cut_off+p_attr_instance.edge_gradient
old_cmp_value = prediction_pair2instance[prediction_pair].recurrence_cut_off + prediction_pair2instance[prediction_pair].edge_gradient
if new_cmp_value>old_cmp_value:
prediction_pair2instance[prediction_pair] = p_attr_instance
counter += 1
if self.report and counter%10000==0:
sys.stderr.write("%s%s/%s"%('\x08'*20, counter, real_counter))
if self.report:
sys.stderr.write("%s%s/%s"%('\x08'*20, counter, real_counter))
del reader
sys.stderr.write("Done.\n")
return prediction_pair2instance
def output_p_gene_id_list(self, curs, schema_instance1, schema_instance2, p_gene_id_list, writer, pic_output_dir,\
pga_instance1, pga_instance2, cluster_info_instance, simple):
"""
10-15-05
add score1 and is_accepted1
10-17-05
score and is_accepted depend on whether pga_instance is None or not
10-17-05 add simple to allow no graph pictures output
also get prediction from schema_instance1 and calculate the score if prediction is available
10-18-05
sort the p_gene_id_list first
"""
#10-15-05 following sentence slightly different from PredictionFilterByClusterSize.py in the trailing edge_gradient
#and d_matrix is a placeholder
sql_sentence1 = "SELECT p.p_gene_id, p.gene_no, p.go_no, p.is_correct, p.is_correct_l1, \
p.is_correct_lca, p.avg_p_value, p.no_of_clusters, p.cluster_array, p.p_value_cut_off, p.recurrence_cut_off, \
p.connectivity_cut_off, p.cluster_size_cut_off, p.unknown_cut_off, p.depth_cut_off, p.mcl_id, p.lca_list, \
p.vertex_gradient, p.edge_gradient, m.vertex_set, s.edge_set, 'd_matrix', 'r' from %s p, %s s, %s m where \
p.mcl_id=s.splat_id and p.mcl_id=m.mcl_id"%(schema_instance1.p_gene_table, \
schema_instance1.splat_table, schema_instance1.mcl_table)
sql_sentence2 = "SELECT p.p_gene_id, p.gene_no, p.go_no, p.is_correct, p.is_correct_l1, \
p.is_correct_lca, p.avg_p_value, p.no_of_clusters, p.cluster_array, p.p_value_cut_off, p.recurrence_cut_off, \
p.connectivity_cut_off, p.cluster_size_cut_off, p.unknown_cut_off, p.depth_cut_off, p.mcl_id, p.lca_list, \
p.vertex_gradient, p.edge_gradient, m.vertex_set, s.edge_set, 'd_matrix', 'r' from %s p, %s s, %s m where \
p.mcl_id=s.splat_id and p.mcl_id=m.mcl_id"%(schema_instance2.p_gene_table, \
schema_instance2.splat_table, schema_instance2.mcl_table)
writer.writerow(['p_gene_id', 'gene_no', 'go_no', 'is_correct_lca', 'p_value', 'recurrence', 'connectivity',\
'cluster_size', 'unknown_ratio', 'mcl_id', 'lca_list', 'edge_gradient', 'score1', 'is_accepted1', 'score2', 'is_accepted2'])
p_gene_id_list.sort()
for p_gene_id in p_gene_id_list:
#sql_sentence1's prediction infomation is not gonna be displayed
curs.execute("%s and p.p_gene_id=%s"%(sql_sentence1, p_gene_id))
rows = curs.fetchall()
if rows:
p_attr_instance1 = prediction_attributes(rows[0], type=3)
else:
p_attr_instance1 = None
#sql_sentence2's prediction infomation is going to be displayed
curs.execute("%s and p.p_gene_id=%s"%(sql_sentence2, p_gene_id))
rows = curs.fetchall()
if rows:
p_attr_instance2 = prediction_attributes(rows[0], type=3)
if pga_instance1 and p_attr_instance1:
(is_accepted1, score1) = pga_instance1.prediction_accepted(p_attr_instance1.go_no, \
[-math.log(p_attr_instance1.p_value_cut_off), p_attr_instance1.recurrence_cut_off, \
p_attr_instance1.connectivity_cut_off, p_attr_instance1.cluster_size_cut_off, \
p_attr_instance1.edge_gradient])
else:
is_accepted1, score1 = None, None
if pga_instance2:
(is_accepted2, score2) = pga_instance2.prediction_accepted(p_attr_instance2.go_no, \
[-math.log(p_attr_instance2.p_value_cut_off), p_attr_instance2.recurrence_cut_off, \
p_attr_instance2.connectivity_cut_off, p_attr_instance2.cluster_size_cut_off, \
p_attr_instance2.edge_gradient])
else:
is_accepted2, score2 = None, None
writer.writerow([p_attr_instance2.p_gene_id, p_attr_instance2.gene_no, p_attr_instance2.go_no, \
p_attr_instance2.is_correct_lca, p_attr_instance2.avg_p_value, p_attr_instance2.recurrence_cut_off,\
p_attr_instance2.connectivity_cut_off, p_attr_instance2.cluster_size_cut_off, p_attr_instance2.unknown_cut_off,\
p_attr_instance2.mcl_id, p_attr_instance2.lca_list, p_attr_instance2.edge_gradient, score1, is_accepted1, \
score2, is_accepted2])
if not simple:
#prepare vertex_set and edge_set to draw graphs
vertex_set = p_attr_instance2.vertex_set[1:-1].split(',')
vertex_set = map(int, vertex_set)
edge_set = p_attr_instance2.edge_set[2:-2].split('},{')
for i in range(len(edge_set)):
edge_set[i] = edge_set[i].split(',')
edge_set[i] = map(int, edge_set[i])
#following copied from GuiAnalyzer.py
subgraph = cluster_info_instance.graph_from_node_edge_set(vertex_set, edge_set)
graphSrcFname = '/tmp/GuiAnalyzer.dot'
graphFname = os.path.join(pic_output_dir, '%s_%s_%s_%s.png'%(p_attr_instance2.p_gene_id, \
p_attr_instance2.gene_no, p_attr_instance2.go_no, p_attr_instance2.mcl_id))
graphSrcF = open(graphSrcFname, 'w')
graphDotOutput(graphSrcF, subgraph, \
self.gene_no2gene_id, self.gene_no2go_no, \
centralnode=p_attr_instance2.gene_no, function=p_attr_instance2.go_no, weighted=0, )
graphSrcF.close()
plot_type_command='neato -Goverlap=false'
commandline = '%s -Tpng %s -o %s'%(plot_type_command, graphSrcFname, graphFname)
system_call(commandline)
def data_fetch(
self,
curs,
schema_instance,
filter_type,
is_correct_type,
no_of_total_genes,
go_no2gene_no_set,
need_cal_hg_p_value=0,
):
"""
11-09-05
1st get the data from p_gene_table and remove redundancy given filter_type
2nd transform the data to three lists
11-10-05 add a chunk of code to get hg p-value(leave one out) for the prediction
mcl_id2vertex_list might blow the memory.(?)
11-19-05
separate predictions totally into known and unknown
2006-10-30, add avg_degree(vertex_gradient) and unknown_cut_off
"""
sys.stderr.write("Fetching data from old p_gene_table...\n")
prediction_pair2instance = {}
curs.execute(
"DECLARE crs CURSOR FOR SELECT p.p_gene_id, p.gene_no, p.go_no, p.is_correct, p.is_correct_l1, \
p.is_correct_lca, p.avg_p_value, p.no_of_clusters, p.cluster_array, p.p_value_cut_off, p.recurrence_cut_off, \
p.connectivity_cut_off, p.cluster_size_cut_off, p.unknown_cut_off, p.depth_cut_off, p.mcl_id, p.lca_list, p.vertex_gradient,\
p.edge_gradient from %s p"
% (schema_instance.p_gene_table)
)
curs.execute("fetch 10000 from crs")
rows = curs.fetchall()
counter = 0
real_counter = 0
while rows:
for row in rows:
p_attr_instance = prediction_attributes(row, type=2)
prediction_pair = (p_attr_instance.gene_no, p_attr_instance.go_no)
if prediction_pair not in prediction_pair2instance:
prediction_pair2instance[prediction_pair] = p_attr_instance
real_counter += 1
else: # remove redundancy
if filter_type == 1:
new_cmp_value = p_attr_instance.recurrence_cut_off
old_cmp_value = prediction_pair2instance[prediction_pair].recurrence_cut_off
elif filter_type == 2:
new_cmp_value = p_attr_instance.edge_gradient
old_cmp_value = prediction_pair2instance[prediction_pair].edge_gradient
elif filter_type == 3:
new_cmp_value = p_attr_instance.recurrence_cut_off + p_attr_instance.edge_gradient
old_cmp_value = (
prediction_pair2instance[prediction_pair].recurrence_cut_off
+ prediction_pair2instance[prediction_pair].edge_gradient
)
if new_cmp_value > old_cmp_value:
prediction_pair2instance[prediction_pair] = p_attr_instance
counter += 1
if self.report:
sys.stderr.write("%s%s/%s" % ("\x08" * 20, counter, real_counter))
curs.execute("fetch 10000 from crs")
rows = curs.fetchall()
unknown_prediction_ls = []
known_prediction_ls = []
unknown_data = [] # 11-19-05
known_data = []
for prediction_pair, p_attr_instance in prediction_pair2instance.iteritems():
# 11-10-05
mcl_id2vertex_list = {}
if need_cal_hg_p_value:
mcl_id = p_attr_instance.mcl_id
if mcl_id not in mcl_id2vertex_list:
mcl_id2vertex_list[mcl_id] = self.get_vertex_list(curs, schema_instance, mcl_id)
p_attr_instance.p_value_cut_off = cal_hg_p_value(
p_attr_instance.gene_no,
p_attr_instance.go_no,
mcl_id2vertex_list[mcl_id],
no_of_total_genes,
go_no2gene_no_set,
r,
)
is_correct = p_attr_instance.is_correct_dict[is_correct_type]
# 2006-10-30, add avg_degree(vertex_gradient) and unknown_cut_off
data_row = [
p_attr_instance.p_value_cut_off,
p_attr_instance.recurrence_cut_off,
p_attr_instance.connectivity_cut_off,
p_attr_instance.cluster_size_cut_off,
p_attr_instance.edge_gradient,
p_attr_instance.vertex_gradient,
p_attr_instance.unknown_cut_off,
p_attr_instance.gene_no,
p_attr_instance.go_no,
is_correct,
]
if is_correct != -1:
known_data.append(data_row) # to do fitting
known_prediction_ls.append(p_attr_instance)
else:
unknown_data.append(data_row)
unknown_prediction_ls.append(p_attr_instance)
sys.stderr.write("Done fetching data.\n")
return unknown_prediction_ls, known_prediction_ls, unknown_data, known_data
