def _5train(f1out,
eachfile,
train,
dirout_feature,
f2resultOut,
fin_model=None):
'''
training the model
'''
print('training on the model')
dir_in = dirout_feature
dirout = os.path.join(f2resultOut, eachfile)
check_path(dirout)
validate = {}
validate['fin_pair'] = os.path.join(f1out, eachfile, 'validate.txt')
validate['dir_in'] = dir_in
onehot = True
entry(
dirout,
train,
dir_in,
model_type=Param.CNN1D_OH,
limit=0,
onehot=onehot,
kernel_size=90,
epochs=80,
# epochs=30,
filters=300,
batch_size=500,
validate=validate,
fin_model=fin_model)
def _1posiPair(dirDIP, dirout):
'''
posi tmp and nontmp
for each
save 1000 pair
save 880 protein fasta file/8DIPPredict/data/Ecoli/2pair.fasta
save 369 tmp file/8DIPPredict/data/Ecoli/2tmp.fasta
save 511 nontmp file/8DIPPredict/data/Ecoli/2nontmp.fasta
# fin = '/home/jjhnenu/data/PPI/release/otherdata/DIP/Ecoli/2Ecoli20170205_id_pair_12246.txt'
# foutdir = 'file/8DIPPredict/data/Ecoli'
# check_path(foutdir)
# handlePair(foutdir,sep='\t',fin=fin,jumpStep=[5],keepOne=True)
'''
# dirDIP = '/home/jjhnenu/data/PPI/release/otherdata/DIP/'
# dirout = 'file/8DIPPredict/data/'
for eachdir in os.listdir(dirDIP):
if eachdir not in ['Ecoli', 'Mus', 'Human', 'SC', 'HP']: continue
currentdir = os.path.join(dirDIP, eachdir)
for eachfile in os.listdir(currentdir):
if 'id_pair' not in eachfile: continue
fin = os.path.join(currentdir, eachfile)
foutdir = os.path.join(dirout, eachdir)
check_path(foutdir)
handlePair(foutdir, sep='\t', fin=fin, jumpStep=[5], keepOne=True)
def composeNegaPair(currentdir, fpositive, foutdir):
ftmp = os.path.join(currentdir, '2tmp.list')
fnontmp = os.path.join(currentdir, '2nontmp.list')
fposi = os.path.join(currentdir, '2pair.tsv')
f1pair = os.path.join(foutdir, '1pair.tsv')
f2pair = os.path.join(foutdir, '2pair.tsv')
f2pairInfo = os.path.join(foutdir, '2pairInfo.tsv')
f3pairInfo = os.path.join(foutdir, '3pairInfo.tsv')
f4pairInfo_subcell = os.path.join(foutdir, '4pairInfo_subcell.tsv')
f4pairInfo_subcell_differ = os.path.join(foutdir,
'4pairInfo_subcell_differ.tsv')
dirout_related = os.path.join(foutdir, '4pairInfo_subcell_differ_related')
check_path(dirout_related)
df = pd.read_table(fposi, header=None)
composeTMP_nonTMP(ftmp, fnontmp, f1pair, int(df.shape[0] * 1.5))
dropPositiveAndRepeate(f1pair, fpositive, f2pair)
dropPositiveAndRepeate(f1pair, fposi, f2pair)
getPairInfo_TMP_nonTMP(f2pair,
f2pairInfo,
sep='\t',
checkTMP=False,
keepOne=True)
saveQualified(f2pairInfo, f3pairInfo)
handleRow(f3pairInfo, f4pairInfo_subcell, calcuSubcell)
saveDifferSubcell(f4pairInfo_subcell, f4pairInfo_subcell_differ)
saveRelated(f4pairInfo_subcell_differ, dirout_related)
def saveRelated(fin_info,dirout):
'''
:param fin_info:
:param dirout:
:return:
# fin_info = os.path.join(dirout, '2subcellular_differ.tsv')
'''
check_path(dirout)
print('save related to',dirout)
fout_fasta = os.path.join(dirout, '2pair.fasta')
fout_tmp_fasta = os.path.join(dirout, '2tmp.fasta')
fout_nontmp_fasta = os.path.join(dirout, '2nontmp.fasta')
f2positive = os.path.join(dirout, '2pair.tsv')
f2tmp = os.path.join(dirout, '2tmp.list')
f2nontmp = os.path.join(dirout, '2nontmp.list')
f2all = os.path.join(dirout, '2all.list')
f2tmp_info = os.path.join(dirout, '2tmp_info.tsv')
f2nontmp_info = os.path.join(dirout, '2nontmp_info.tsv')
f2all_info = os.path.join(dirout, '2all_info.tsv')
simplifyTable(fin_info, f2positive)
extractPairAndFasta(fin_info, fout_fasta, fout_tmp_fasta=fout_tmp_fasta,
fout_nontmp_fasta=fout_nontmp_fasta)
getproteinlist(fin_info,
ftmp=f2tmp, fnontmp=f2nontmp, fall=f2all,
ftmp_info=f2tmp_info, ftmp_nontmp_info=f2nontmp_info, fall_info=f2all_info)
def _2_1combineFasta(fposiInfo,fnegaInfo,dirout):
'''
:param fposiInfo:
:param fnegaInfo:
:param dirout:
:return:
fin_pair = '%s/dirRelated/2pair.tsv'%dirout
fasta = '%s/dirRelated/2pair.fasta'%dirout
'''
'''
config path
'''
# fposiInfo = 'file/10humanTrain/1positive/9human_human.tsv'
# fnegaInfo = 'file/10humanTrain/2negative/4subcellular/2subcellular_differ.tsv'
#
# dirout = 'file/10humanTrain/3cluster/all'
check_path(dirout)
fpairinfo = os.path.join(dirout, '1pairinfo.tsv')
dirRelated = os.path.join(dirout, 'dirRelated')
'''
concat positive and negative with info
'''
df1 = pd.read_table(fposiInfo,header=None)
df2 = pd.read_table(fnegaInfo,header=None)
df3 = pd.concat([df1,df2])
df3.to_csv(fpairinfo,header=None,index=None,sep='\t')
'''
save 19724 protein fasta file/10humanTrain/3cluster/all/dirRelated/2pair.fasta
save 5089 tmp file/10humanTrain/3cluster/all/dirRelated/2tmp.fasta
save 14635 nontmp file/10humanTrain/3cluster/all/dirRelated/2nontmp.fasta
'''
saveRelated(fpairinfo, dirRelated)
def groupCalculate(dirin, filetype='all'):
"""
/home/jjhnenu/data/PPI/release/result/group/p_fp_1_1/1/all/_evaluate.txt
:param dirin: /home/jjhnenu/data/PPI/release/result/group
:return:
"""
for eachdir in os.listdir(dirin):
subdir = os.path.join(
dirin,
eachdir) # /home/jjhnenu/data/PPI/release/result/group/p_fp_1_1/
data = []
columns = ['Loss', 'Acc', 'Precision', 'Recall', 'F1score', 'MCC']
# columns = ['loss', 'acc', 'metric_precision', 'metric_recall', 'metric_F1score', 'matthews_correlation']
print(columns)
for eachsubdir in os.listdir(subdir): # 0 1 2 3 4 5
fin = os.path.join(subdir, eachsubdir, filetype, '_evaluate.txt')
# fin = os.path.join(subdir,eachsubdir,filetype,'_history_dict.txt')
if not os.access(fin, os.F_OK): continue
with open(fin, 'r') as fi:
line = fi.readline()[:-1]
# sum += np.array(line.split(':')[-1][1:-1].split(','))
data.append(line.split(':')[-1][1:-1].split(','))
print(str(line.split(':')[-1][1:-1].split(','))[1:-1])
mydata = pd.DataFrame(data)
t = mydata.apply(pd.to_numeric)
t.columns = columns
t.loc['mean'] = t.apply(lambda x: x.mean())
dirout = os.path.join(subdir.replace('result', 'statistic'))
check_path(dirout)
# float_format = '%.3f'
t.sort_index(inplace=True)
t.to_csv(os.path.join(dirout, 'result.csv'), index=True, header=True)
print(dirout)
def load_ed_models(model_paths, pc):
ed_models = []
ed_model_params = []
for i, path in enumerate(model_paths):
print '...Loading nmt model {}'.format(i)
ed_model_folder = check_path(path,
'ED_MODEL_FOLDER_{}'.format(i),
is_data_path=False)
best_model_path = ed_model_folder + '/bestmodel.txt'
hypoparams_file_reader = codecs.open(ed_model_folder + '/best.dev',
'r', 'utf-8')
hyperparams_dict = dict([
line.strip().split(' = ')
for line in hypoparams_file_reader.readlines()
])
model_hyperparams = {
'INPUT_DIM': int(hyperparams_dict['INPUT_DIM']),
'HIDDEN_DIM': int(hyperparams_dict['HIDDEN_DIM']),
'LAYERS': int(hyperparams_dict['LAYERS']),
'VOCAB_PATH': hyperparams_dict['VOCAB_PATH']
}
# a fix for vocab path when transferring files b/n vm
model_hyperparams['VOCAB_PATH'] = check_path(path + '/vocab.txt',
'vocab_path',
is_data_path=False)
ed_model_params.append(pc.add_subcollection('ed{}'.format(i)))
ed_model = SoftAttention(ed_model_params[i], model_hyperparams,
best_model_path)
ed_models.append(ed_model)
return ed_models
def _7trainAndTest(dirout_feature, fin_train, fin_validate, dirout):
# time 664909.4274818897 ~ 7.6 day
'''
training the model
'''
print('training on the model')
check_path(dirout)
validate = {}
validate['fin_pair'] = fin_validate
validate['dir_in'] = dirout_feature
onehot = True
entry(
dirout,
fin_train,
dirout_feature,
model_type=Param.CNN1D_OH,
limit=0,
onehot=onehot,
kernel_size=90,
epochs=80,
# epochs=30,
filters=300,
batch_size=50,
validate=validate)
'''
testing on the model
'''
print('testing the model')
def calculateResults(dirout,
dirin,
filename='_evaluate.txt',
row=0,
resultfilename='result.csv'):
"""
%s\%s\_evaluate.txt
:param dirin: contains a list of \%s\_evaluate.txt
:return:
dirin = '/home/19jiangjh/data/PPI/release/result_in_paper/alter_ratio/p_fw_v1_train_validate_v2_fixpositive_2/2/test_DIP/'
dirout = dirin
calculateResults(dirout,dirin,filename='log.txt',row = 2,resultfilename = 'result.csv')
"""
# dirin = r'E:\githubCode\BioComputeCodeStore\JiangJiuhong\data\PPI\stage2\processPair2445\pair\positiveV1\onehot\result'
check_path(dirout)
count = 0
data = []
# columns = ['loss', 'acc', 'metric_precision', 'metric_recall', 'metric_F1score', 'matthews_correlation']
columns = ['Loss', 'Acc', 'Precision', 'Recall', 'F1score', 'MCC']
indexs = []
print(columns)
print(dirin)
for eachdir in os.listdir(dirin):
print(eachdir)
if '.' in eachdir: continue
fin = os.path.join(dirin, eachdir)
sep = '\\' if '\\' in filename else '/'
if sep in filename:
for f in filename.split(sep):
fin = os.path.join(fin, f)
else:
fin = os.path.join(fin, filename)
# fin = '%s\%s\_evaluate.txt' % (dirin, eachdir)
if not os.access(fin, os.F_OK):
print('not access to:', fin)
continue
with open(fin, 'r') as fi:
real_row = 0
while (real_row != row):
fi.readline()
real_row = real_row + 1
line = fi.readline()[:-1]
# sum += np.array(line.split(':')[-1][1:-1].split(','))
line = line.replace('nan', '0')
print('****************', line, '********************')
data.append(line.split(':')[-1][1:-1].split(','))
indexs.append(eachdir)
count = count + 1
print(str(line.split(':')[-1][1:-1].split(','))[1:-1])
mydata = pd.DataFrame(data)
mydata.replace('nan', 0, inplace=True)
t = mydata.apply(pd.to_numeric)
t.loc['mean'] = t.apply(lambda x: x.mean())
indexs.append('mean')
t.index = indexs
t.columns = columns
t.sort_index(inplace=True)
t.to_csv(os.path.join(dirout, resultfilename), index=True, header=True)
def getNpy(self, fin_fasta, out_dir, multi=True):
check_path(out_dir)
for ID, seq in self.getYield(fin_fasta, multi=multi):
if p.checkProtein(seq, 50, 2000, uncomm=True):
filename = os.path.join(out_dir, "%s.npy" % ID)
result = self.seq2num(seq)
if len(result) != 0:
np.save(filename, result)
def getPhsi_Blos(self, fin_fasta, out_dir, multi=True, checkprotein=True):
check_path(out_dir)
p = Protein()
for ID, seq in self.getYield(fin_fasta, multi=multi):
if not checkprotein or p.checkProtein(seq, 50, 2000, uncomm=True):
filename = os.path.join(out_dir, "%s.npy" % ID)
result = self.phsi_blos(seq)
if len(result) != 0:
np.save(filename, result)
def savepredict(fin_pair, dir_in, fin_model, dirout_result):
# fin_pair = '/home/19jiangjh/data/PPI/release/pairdata/p_fw/1/0/test.txt'
# dir_in = '/home/19jiangjh/data/PPI/release/feature/p_fp_fw_19471'
# fin_model = '/home/19jiangjh/data/PPI/release/result_in_paper/alter_ratio/p_fw_train_validate/1/_my_model.h5'
# dirout_result = '/home/19jiangjh/data/PPI/release/result_in_paper/alter_ratio/p_fw_train_validate/1/test'
check_path(dirout_result)
onehot = True
dataarray = BaseData().loadTest(fin_pair,
dir_in,
onehot=onehot,
is_shuffle=False)
x_test, y_test = dataarray
model = models.load_model(fin_model, custom_objects=MyEvaluate.metric_json)
result = model.evaluate(x_test, y_test, verbose=False, batch_size=90)
result_predict = model.predict(x_test, batch_size=90)
result_predict = result_predict.reshape(-1)
result_class = model.predict_classes(x_test, batch_size=90)
result_class = result_class.reshape(-1)
y_test = y_test.reshape(-1)
print('Loss:%f,ACC:%f' % (result[0], result[1]))
df = pd.read_table(fin_pair, header=None)
# df.columns = ['tmp', 'nontmp']
df.rename(columns={0: 'tmp', 1: 'nontmp'}, inplace=True)
df['real_label'] = list(y_test)
df['predict_label'] = result_class
df['predict'] = result_predict
df.to_csv(os.path.join(dirout_result, 'result.csv'), index=False)
result_manual = MyEvaluate().evaluate_manual(y_test, result_predict)
print(
'[acc,metric_precision, metric_recall, metric_F1score, matthews_correlation]'
)
print(result_manual)
print('[acc,precision,sensitivity,f1,mcc,aps,aucResults,specificity]')
result_manual2 = calculate_performance(len(x_test), y_test, result_class,
result_predict)
print(result_manual2)
with open(os.path.join(dirout_result, 'log.txt'), 'w') as fo:
fo.write('test dataset %s\n' % fin_pair)
fo.write('Loss:%f,ACC:%f\n' % (result[0], result[1]))
fo.write('evaluate result:' + str(result) + '\n')
fo.write(
'manual result:[acc,metric_precision, metric_recall, metric_F1score, matthews_correlation]\n'
)
fo.write('manual result:' + str(result_manual) + '\n')
fo.write(
'manual result2:[acc,precision,sensitivity,f1,mcc,aps,aucResults,specificity]\n'
)
fo.write('manual result2:' + str(result_manual2) + '\n')
fo.flush()
def entry(dirout,
fin_pair,
dir_in,
model_type=Param.CNN1D,
limit=0,
onehot=False,
kernel_size=3,
epochs=60,
filters=250,
batch_size=100,
validate=None):
# model_type = Param.CNN1D
# fin_pair = '/home/jjhnenu/data/PPI/release/data/p_fp_fw_2_1_1/all.txt'
# dir_in = '/home/jjhnenu/data/PPI/release/feature/pssm400_feature_1D/p_fp_fw_2_1_1/all/'
# fout = '/home/jjhnenu/data/PPI/release/result/pssm400_feature_1D/p_fp_fw_2_1_1/all/'
# if 'all' not in dirout:return
check_path(dirout)
print('dirout:', dirout)
# dir_in = dirout.replace(des, 'feature')
print('dir_in:', dir_in)
print('fin_pair', fin_pair)
bd = BaseData()
if validate == None:
(x_train, y_train), (x_test, y_test) = bd.load(fin_pair,
dir_in,
test_size=0.1,
limit=limit,
onehot=onehot)
else:
x_test, y_test = bd.loadTest(validate['fin_pair'],
validate['dir_in'],
onehot=onehot,
is_shuffle=True)
x_train, y_train = bd.loadTest(fin_pair,
dir_in,
onehot=onehot,
is_shuffle=True)
print('Build and fit model...')
print('x_train.shape[1:]', x_train.shape[1:])
mm = MyModel(
model_type=model_type,
input_shape=x_train.shape[1:],
filters=filters,
kernel_size=kernel_size,
pool_size=2,
hidden_dims=250,
batch_size=batch_size,
epochs=epochs,
)
mm.process(dirout, x_train, y_train, x_test, y_test)
print('save result to %s' % dirout)
def base_compose(self,
dirout_feature,
fin_pair,
dir_feature_db,
feature_type='V_PSSM',
fout_pair=''):
check_path(dirout_feature)
fo = open(fout_pair, 'w') if fout_pair != '' else None
row = 0
for pairs in getPairs(fin_pair):
a = pairs[0]
b = pairs[1]
# print(pairs) # ['O35668', 'P00516']
fa = os.path.join(dir_feature_db, a + '.npy')
fb = os.path.join(dir_feature_db, b + '.npy')
row = row + 1
print('loading %d th feature pair' % row)
if not (os.access(fa, os.F_OK) and os.access(fb, os.F_OK)):
print(
'===============features of pairs not found %s %s================'
% (a, b), os.access(fa, os.F_OK), os.access(fb, os.F_OK))
continue
pa = np.load(fa, allow_pickle=True)
pb = np.load(fb, allow_pickle=True)
if (len(pa) < 50 or len(pa) > 2000
or max(pa) > 20) or (len(pb) < 50 or len(pb) > 2000
or max(pb) > 20):
print('wrong length or x')
continue
if fo != None:
fo.write('%s\t%s\n' % (a, b))
fo.flush()
# padding
if feature_type == Feature_type.V_PSSM:
pc = self.padding_PSSM(pa, pb, vstack=True)
elif feature_type == Feature_type.H_PSSM:
pc = self.padding_PSSM(pa, pb, vstack=False)
elif feature_type == Feature_type.SEQ_1D:
pc = self.padding_seq1D(pa, pb, vstack=False)
# elif feature_type == Feature_type.SEQ_1D_OH:pc = self.padding_seq1D(pa,pb,vstack=False)
elif feature_type == Feature_type.SEQ_2D:
pc = self.padding_seq2D(pa, pb)
else:
print('incoreect feature_type')
return
# 保存padding后的成对特征
fout = os.path.join(dirout_feature, "%s_%s.npy" % (a, b))
np.save(fout, pc)
del pc, pa, pb
if fo != None:
fo.close()
def build_img(self,
font_name='',
bg_color=None,
b_color_img_name="",
txt_freq=None):
try:
if not font_name.strip():
pass
else:
self.font_name = font_name
self.font_path = os.path.join(my_dirpath,
"../font/%s.ttf" % self.font_name)
self.bg_color = bg_color
self.back_color_img = imread(
os.path.join(my_dirpath, "../image/%s" % b_color_img_name))
my_log.logger.info("Start creating word clouds...")
wc = WordCloud(
font_path=self.font_path, #设置字体
mode="RGBA",
background_color=self.bg_color, #背景颜色
max_words=self.max_words, # 词云显示的最大词数
mask=self.back_color_img, #设置背景图片
max_font_size=self.max_font_size, #字体最大值
random_state=self.random_state, #设置有多少种随机生成状态，即有多少种配色方案
colormap="viridis", #随机颜色
relative_scaling=1,
scale=1.2)
wc.generate_from_frequencies(txt_freq)
#根据图片的颜色布局进行着色
# image_colors = ImageColorGenerator(self.back_color_img)
# wc.recolor(color_func=image_colors)
# 根据给定的颜色值进行渲染
# grouped_color_func = GroupedColorFunc(color_to_words, default_color)
# wc.recolor(color_func=grouped_color_func)
# 绘制词云,保存图片
save_img_path = os.path.join(my_dirpath, "./image/build_img/")
check_path(save_img_path)
img_path = save_img_path + str(time.time()) + ".png"
wc.to_file(img_path)
my_log.logger.info("build img success.img_path:%s" % img_path)
except Exception as e:
my_log.logger.error("build img error...")
my_log.logger.error(e)
def _3clusters(fin_posi,fin_nega,fin_tmp,fin_nontmp,dirout):
'''
:param fin_posi:
:param fin_nega:
:param fin_tmp:
:param fin_nontmp:
:param dirout:
:return:
f4posi 'file/10humanTrain/3cluster/4posi.tsv'
f4nega 'file/10humanTrain/3cluster/4nega.tsv'
cd hit 0.4 cd-hit tool :http://weizhong-lab.ucsd.edu/cdhit_suite/cgi-bin/index.cgi?cmd=cd-hit
get *.clstr file
'''
# fin_posi = 'file/10humanTrain/1positive/2pair.tsv' # (44210, 2)
# fin_nega = 'file/10humanTrain/2negative/4subcellular/dirRelated/2pair.tsv' # (61323, 2)
# fin_tmp = 'file/10humanTrain/3cluster/1tmp.clstr'
# fin_nontmp = 'file/10humanTrain/3cluster/1nontmp.clstr'
# dirout = 'file/10humanTrain/3cluster/'
check_path(dirout)
f2pair = os.path.join(dirout,'1pair.tsv')
f3out_tmp = os.path.join(dirout,'3tmp.tsv')
f3out_nontmp = os.path.join(dirout,'3nontmp.tsv')
f3pair = os.path.join(dirout,'3pair.tsv')
f3pair_clstr = os.path.join(dirout,'3pair_clstr.tsv')
f4pair = os.path.join(dirout,'4pair.tsv')
f4posi = os.path.join(dirout,'4posi.tsv')
f4nega = os.path.join(dirout,'4nega.tsv') # (60697, 3)
'''
concat positive and negative pair
'''
concatPAN(fin_posi, fin_nega, f2pair)
cluster2Table(fin_tmp,f3out_tmp)
cluster2Table(fin_nontmp, f3out_nontmp)
pairWithClusterLable(f2pair,f3out_tmp,f3out_nontmp,fout_clus=f3pair_clstr,fout=f3pair)
'''
extract positive,negative
'''
saveRelated_posi_nega(f3pair,f4pair,f4posi,f4nega)
def _1posiSampleHumanPair(fin_1posiInfo,dirout):
'''
positive samples human-human
'''
# fin_1posiInfo = 'file/1positive/2tmp_nontmp_info_qualified.tsv'
# dirout = 'file/10humanTrain/1positive'
check_path(dirout)
f8tmp_species = os.path.join(dirout, '8tmp_species.tsv')
f8nontmp_species = os.path.join(dirout, '8nontmp_species.tsv')
f8species = os.path.join(dirout, '8species.tsv')
f8sameSpecies = os.path.join(dirout, '8sameSpecies.tsv')
f8posiSpecies = os.path.join(dirout, '8posiSpecies.tsv')
f9human_related = os.path.join(dirout, '9human_related.tsv')
f9human_human = os.path.join(dirout, '9human_human.tsv') # 44210 pairinfo
findSpecies(fin_1posiInfo,f8species, f8tmp_species, f8nontmp_species, f8sameSpecies,f8posiSpecies,col=[1,8])
species = 'HUMAN'
relatedSpecies(f8posiSpecies, species, f9human_related, f9human_human, col=[0,7,14,15])
fout_fasta = os.path.join(dirout, '2pair.fasta')
fout_tmp_fasta = os.path.join(dirout, '2tmp.fasta')
fout_nontmp_fasta = os.path.join(dirout, '2nontmp.fasta')
f2positive = os.path.join(dirout, '2pair.tsv')
f2tmp = os.path.join(dirout, '2tmp.list')
f2nontmp = os.path.join(dirout, '2nontmp.list')
f2all = os.path.join(dirout, '2all.list')
f2tmp_info = os.path.join(dirout, '2tmp_info.tsv')
f2nontmp_info = os.path.join(dirout, '2nontmp_info.tsv')
f2all_info = os.path.join(dirout, '2all_info.tsv')
# save 11995 protein fasta file/10humanTrain/1positive/2pair.fasta
# save 3513 tmp file/10humanTrain/1positive/2tmp.fasta
# save 8482 nontmp file/10humanTrain/1positive/2nontmp.fasta
simplifyTable(f9human_human, f2positive)
extractPairAndFasta(f9human_human, fout_fasta, fout_tmp_fasta=fout_tmp_fasta,
fout_nontmp_fasta=fout_nontmp_fasta)
getproteinlist(f9human_human,
ftmp=f2tmp, fnontmp=f2nontmp, fall=f2all,
ftmp_info=f2tmp_info, ftmp_nontmp_info=f2nontmp_info, fall_info=f2all_info)
def save(self,
dirout,
flist,
ratios,
limit,
labels=None,
sep='\t',
filename='all.txt',
groupcount=1,
repeate=True):
"""
same length of flist,ratios,labels if not None
:param fout:
:param flist:
:param ratios:
:param limit:
:param labels:[1,0,0] 样本中无标签时，需要提供这个标签
:param sep:
:return:
case
dirin = r'E:\githubCode\BioComputeCodeStore\JiangJiuhong\data\PPI\release\data'
fin_p = r'%s\positive_2049.txt' % dirin
fin_fp = r'%s\negative_fpositive_10245.txt' % dirin
fin_fw = r'%s\negative_fswissprot_7781.txt' % dirin
fout = r'%s\p_fp_fw_2_1_1\all.txt' % dirin
flist = [fin_p, fin_fp,fin_fw]
ratios = [0.5,0.25,0.25]
labels = [1,0,0]
limit = 2049
ComposeData().save(fout, flist, ratios, limit, labels=labels)
"""
for idx, elem in enumerate(
self.doCompose(flist,
ratios,
limit,
labels=labels,
groupcount=groupcount,
repeate=repeate)):
data = pd.concat(elem)
data = data.sample(frac=1)
fout = os.path.join(dirout, str(idx))
check_path(fout)
fout = os.path.join(fout, filename)
quick_save(data, fout, sep=sep)
def save(self):
for i in range(len(self.fin)):
dirout = os.path.join(self.dirin, self.type[i])
check_path(dirout)
ComposeData().save(dirout,
self.fin[i],
self.ratio[i],
self.limit,
labels=self.label[i],
filename='all.txt')
print('divided to train and test')
for eachdir in os.listdir(dirout):
if '.' in eachdir: continue
fin = os.path.join(dirout, eachdir, 'all.txt')
train = os.path.join(dirout, eachdir, 'train.txt')
test = os.path.join(dirout, eachdir, 'test.txt')
fouts = [train, test]
ratios = [0.8, 0.2]
PairDealer().part(fin, ratios, fouts)
def process(fin_fasta, dir_feature_db, dir_feature, dir_pair):
'''
generate featuredb and feature(feature pair)
:param fin_fasta:>ID\nseq\n
:param dir_feature_db:
:param dir_feature:
:param dir_pair: contains several protein file
# ID pair proteinA\tproteinB\n
:return:
'''
check_path(dir_feature_db)
check_path(dir_feature)
# fasta to feature
fd = FastaDealer()
fd.getNpy(fin_fasta, dir_feature_db)
for eachfile in os.listdir(dir_pair):
print(eachfile)
fin_pair = os.path.join(dir_pair, eachfile)
BaseFeature().base_compose(dir_feature,
fin_pair,
dir_feature_db,
feature_type=Feature_type.SEQ_1D)
def get_city(self):
"""
获取城市列表
"""
if check_path(self.city_names):
try:
with pathlib.Path(self.city_names).open("r") as city_file:
self.citys = json.loads(city_file.read())
return
except Exception as e:
error(f"城市文件读取错误: {e}")
try:
resp = self.menu_session.get(self.url_list["citys"])
self.save_debug_file(resp.text, "get_city.html")
res = self.reg_list["citys"].findall(resp.text)
addsucess()
self.citys = json.loads(res[0])
create_json(self.citys, self.city_names)
except Exception as e:
error(f"获取城市信息失败: {e}")
addfailed()
exit()
success(f"{len(self.citys.keys())} citys")
def savepredict(fin_pair,
dir_in,
fin_model,
dirout_result,
batch_size=90,
limit=0,
posi=False,
onehot=True):
# fin_pair = '/home/19jiangjh/data/PPI/release/pairdata/p_fw/1/0/test.txt'
# dir_in = '/home/19jiangjh/data/PPI/release/feature/p_fp_fw_19471'
# fin_model = '/home/19jiangjh/data/PPI/release/result_in_paper/alter_ratio/p_fw_train_validate/1/_my_model.h5'
# dirout_result = '/home/19jiangjh/data/PPI/release/result_in_paper/alter_ratio/p_fw_train_validate/1/test'
check_path(dirout_result)
print('predict ', fin_pair, '...')
print('using feature ', dir_in, '...')
print('save result in ', dirout_result)
df = pd.read_table(fin_pair, header=None)
if df.shape[1] != 3:
df[2] = 1 if posi else 0
dataarray = BaseData().loadTest(fin_pair,
dir_in,
onehot=onehot,
is_shuffle=False,
limit=limit)
x_test, y_test = dataarray
print('load model...', fin_model)
# from tensorflow.keras.models import load_model
# model = load_model(fin_model, custom_objects=MyEvaluate.metric_json)
model = models.load_model(fin_model, custom_objects=MyEvaluate.metric_json)
model.compile(loss='binary_crossentropy',
optimizer='adam',
metrics=MyEvaluate.metric)
result = model.evaluate(x_test, y_test, verbose=1, batch_size=batch_size)
result_predict = model.predict(x_test, batch_size=batch_size)
result_class = (result_predict > 0.5).astype("int32")
result_predict = result_predict.reshape(-1)
# result_class = model.predict_classes(x_test,batch_size=batch_size)
# UserWarning: `model.predict_classes()` is deprecated and will be removed after 2021-01-01.
result_class = result_class.reshape(-1)
# y_test = y_test.reshape(-1)
print('Loss:%f,ACC:%f' % (result[0], result[1]))
if limit != 0: df = df[:limit]
df.columns = ['tmp', 'nontmp', 'real_label']
# df.rename(columns={0: 'tmp', 1: 'nontmp'}, inplace=True)
# df['real_label'] = list(y_test)
df['predict_label'] = result_class
df['predict'] = result_predict
df.sort_values(by=['predict'], ascending=False).to_csv(os.path.join(
dirout_result, 'result.csv'),
index=False)
# result_manual = MyEvaluate().evaluate_manual(y_test, result_predict)
# print('[acc,metric_precision, metric_recall, metric_F1score, matthews_correlation]')
# print(result_manual)
# print('[acc,precision,sensitivity,f1,mcc,aps,aucResults,specificity]')
# result_manual2 =calculate_performance(len(x_test), y_test, result_class, result_predict)
# print(result_manual2)
with open(os.path.join(dirout_result, 'log.txt'), 'w') as fo:
fo.write('test dataset %s\n' % fin_pair)
fo.write('Loss:%f,ACC:%f\n' % (result[0], result[1]))
fo.write('evaluate result:' + str(result) + '\n')
fo.write(
'manual result:[acc,metric_precision, metric_recall, metric_F1score, matthews_correlation]\n'
)
# fo.write('manual result:' + str(result_manual) + '\n')
# fo.write('manual result2:[acc,precision,sensitivity,f1,mcc,aps,aucResults,specificity]\n')
# fo.write('manual result2:'+str(result_manual2)+'\n')
fo.flush()
projection = {'_id': True, 'UNIPROID':True}
docs = do.GetALL(projection=projection,limit=0)
for protein in docs:
if protein['UNIPROID'] == '':continue
for pro in multiSplit(protein['UNIPROID']):
yield pro.strip()
def writeProtins(fout):
proteins = []
for protein in getallProtein():
proteins.extend(protein)
saveList(getallProtein(),fout)
if __name__ == '__main__':
print('start', time.strftime('%Y-%m-%d %H:%M:%S', time.localtime()))
start = time.time()
db_name = 'ttd'
table_target = 'target'
fout = 'file/otherfile/2ttd_target_protein.list'
check_path('file/otherfile/')
writeProtins(fout)
pass
print('stop', time.strftime('%Y-%m-%d %H:%M:%S', time.localtime()))
print('time', time.time() - start)
f2nontmpInfo = os.path.join(dirout,'2nontmpInfo.tsv')
f2tmp = os.path.join(dirout, '2tmp.list')
f2nontmp = os.path.join(dirout, '2nontmp.list')
f3pair = os.path.join(dirout,'3pair.tsv')
f3pair_norepeat = os.path.join(dirout,'3pair_norepeat.tsv')
f3pairInfo = os.path.join(dirout,'2pairInfo.tsv')
f3tmp_fasta = os.path.join(dirout,'3tmp.fasta')
f3nontmp_fasta = os.path.join(dirout,'3nontmp.fasta')
f3all_fasta = os.path.join(dirout,'3all.fasta')
dir3bathData = os.path.join(dirout,'3batchData')
check_path(dir3bathData)
f4sample1k = os.path.join(dir3bathData,'3batchData')
'''
get TMP, nonTMP list
query 5208 tmp and 15186 nontmp
'''
# generateHumanLists(f1tmp, f1nontmp)
'''
get qualified single info
'''
# getSingleInfo(f1tmp,f2tmpInfo,fin_type='single')
# getSingleInfo(f1nontmp,f2nontmpInfo,fin_type='single')
'''
get qualified protein
'''
input file :
positive pair and negative pair
AC-pair of TMP-nonTMP
fasta
generate feature
fasta
'''
print()
'''
fasta to feature
'''
# >ID\nseq\n
fin_fasta = '/home/jjhnenu/data/PPI/release/otherdata/DIP/_DIP_fasta20170301_simple.seq'
dir_feature_db = '/home/jjhnenu/data/PPI/release/otherdata/DIP/_featuredb'
check_path(dir_feature_db)
fd = FastaDealer()
fd.getNpy(fin_fasta, dir_feature_db)
'''
generate feature pair
50-2000 no X
'''
dir_feature_db = '/home/jjhnenu/data/PPI/release/otherdata/DIP/_featuredb'
dir_feature = '/home/jjhnenu/data/PPI/release/otherdata/DIP/_feature/'
# ID pair proteinA\tproteinB\n
dir_pair = '/home/jjhnenu/data/PPI/release/otherdata/DIP/_TMP_nonTMP/'
dirout_pair = '/home/jjhnenu/data/PPI/release/otherdata/DIP/_TMP_nonTMP_qualified/'
check_path(dir_feature)
check_path(dirout_pair)
for eachfile in os.listdir(dir_pair):
print(eachfile)
def crosshumanTrain(modelreuse=False):
f2all = 'file/10humanTrain/3cluster/4pair.tsv' # 正负样本非1：1了
f1out = 'file/10humanTrain/4train/group'
f2out = 'file/10humanTrain/4train/cross/group'
f3out = 'file/10humanTrain/4train/cross'
# dirout_feature = '/home/19jjhnenu/Data/SeqTMPPI2W/feature/129878/'
# f2resultOut = '/home/19jjhnenu/Data/SeqTMPPI2W/result/10humanTrain_80epoch/group_reusemodel_5CV'
dirout_feature = '/root/19jjhnenu/Data/SeqTMPPI2W/feature/129878/'
f2resultOut = '/root/19jjhnenu/Data/SeqTMPPI2W/result/10humanTrain_80epoch/group_reusemodel_5CV'
check_path(f2resultOut)
check_path(f2out)
# flist = [os.path.join(f1out,x,'all.txt') for x in os.listdir(f1out)]
# concatFile(flist,f2all)
'''
train:test = 5:1
'''
train = os.path.join(f3out, 'train_vali.txt')
test = os.path.join(f3out, 'test.txt')
ratios_tvt = [5, 1]
f3outs = [train, test]
# PairDealer().part(os.path.join(f3out,'all.txt'),ratios_tvt,f3outs)
# PairDealer().part(f2all,ratios_tvt,f3outs)
'''
5cv
'''
ratios_tvt = [1] * 5
f3outs = [os.path.join(f2out, '%d.txt' % x) for x in range(5)]
# PairDealer().part(os.path.join(f3out,'all.txt'),ratios_tvt,f3outs)
# PairDealer().part(os.path.join(f3out,'train_vali.txt'),ratios_tvt,f3outs)
'''
cross train
'''
for cv in range(5):
# oldfile = '-1'
oldfile = '2'
for elem in range(5):
if elem == cv: continue
f2dirout = os.path.join(f2resultOut, str(cv))
f3dirout = os.path.join(f2dirout, str(elem))
fin_model = os.path.join(f2dirout, oldfile, '_my_model.h5')
if not os.access(fin_model, os.F_OK) or not modelreuse:
fin_model = None
train = os.path.join(f2out, '%d.txt' % elem)
validate = {}
validate['fin_pair'] = os.path.join(f2out, '%d.txt' % cv)
validate['dir_in'] = dirout_feature
onehot = True
entry(
f3dirout,
train,
dirout_feature,
model_type=Param.CNN1D_OH,
limit=0,
onehot=onehot,
kernel_size=90,
epochs=80,
# epochs=2,
filters=300,
batch_size=500,
validate=validate,
fin_model=fin_model)
oldfile = str(elem)
def crossTrain(f2out, dirout_feature, f2resultOut, modelreuse=True):
'''
cross train and test
'''
f1out = 'file/4train/'
f2outdir = os.path.join(f1out, str(0))
fin_pair = os.path.join(f2outdir, 'all.txt')
f2outdir = os.path.join(f1out, '5CV', 'data')
check_path(f2outdir)
train = os.path.join(f2outdir, 'train_vali.txt')
test = os.path.join(f2outdir, 'test.txt')
ratios_tvt = [5, 1]
f3outs = [train, test]
# PairDealer().part(fin_pair,ratios_tvt,f3outs)
'''
train model
'''
f2outdir = os.path.join(f1out, '5CV', 'data')
check_path(f2outdir)
train = os.path.join(f2outdir, 'train_vali.txt')
# f2out = 'file/4train/5CV/elem'
ratios_tvt = [1] * 5
f3outs = [os.path.join(f2out, '%d.txt' % x) for x in range(5)]
# PairDealer().part(train,ratios_tvt,f3outs)
limit = 0
# eachdir = 'benchmark'
# dirout_feature = '/home/19jjhnenu/Data/Phsi_Blos/feature/%s/' % eachdir
# f2resultOut = '/home/19jjhnenu/Data/SeqTMPPI2W/result/5CV_1'
'''
cross train
'''
for cv in range(5):
oldfile = '-1'
f2dirout = os.path.join(f2resultOut, str(cv))
fin_model = ''
f3dirout = ''
for elem in range(5):
if cv == elem: continue
f3dirout = os.path.join(f2dirout, str(elem))
fin_model = os.path.join(f2dirout, oldfile, '_my_model.h5')
if not os.access(fin_model, os.F_OK) or not modelreuse:
fin_model = None
train = os.path.join(f2out, '%d.txt' % elem)
validate = {}
validate['fin_pair'] = os.path.join(f2out, '%d.txt' % cv)
validate['dir_in'] = dirout_feature
onehot = False
# entry(f3dirout, train, dirout_feature, model_type=Param.CNN1D_OH, limit=0, onehot=onehot, kernel_size=90,
# epochs=80,
# # epochs=2,
# filters=300, batch_size=500, validate=validate,
# fin_model=fin_model)
# entry(f3dirout, train, dirout_feature, model_type=Param.TRANSFORMER, limit=10, onehot=onehot, kernel_size=90,
# # epochs=80,
# epochs=2,
# filters=300, batch_size=500, validate=validate,
# fin_model=fin_model)
#
#
oldfile = str(elem)
# # print(f3dirout)
# calculateResults(f2dirout, f2dirout, resultfilename='result.csv')
# eachdir = 'benchmark'
# eachdir = 'benchmark'
# dirout_feature = '/home/19jjhnenu/Data/Phsi_Blos/feature/%s/' % eachdir
# print('testing the model on test dataset')
# fin_model = os.path.join(f2dirout, oldfile, '_my_model.h5')
# fin_test = 'file/4train/5CV/data/test.txt'
# dirout_result_test = '/home/19jjhnenu/Data/Phsi_Blos/result/5CV_1_test/%d'%cv
# check_path(dirout_result_test)
# savepredict(fin_test, dirout_feature, fin_model, dirout_result_test,batch_size=500,limit=2000,onehot = onehot)
'''
testing on DIP all.txt in DIP/predict
'''
# fin_test = 'file/8DIPPredict/predict/all.txt'
# dirout_feature = '/home/19jjhnenu/Data/Phsi_Blos/feature/%s/' % 'DIP'
# fin_model = os.path.join(f2dirout, oldfile, '_my_model.h5')
# dirout_result_test = '/home/19jjhnenu/Data/Phsi_Blos/result/5CV_1_DIP/%d' % cv
# check_path(dirout_result_test)
# savepredict(fin_test, dirout_feature, fin_model, dirout_result_test, batch_size=500,onehot = onehot)
'''
testing on DIP all.txt in DIP/data
'''
# fin_test = 'file/8DIPPredict/data/all.txt'
# dirout_feature = '/home/19jjhnenu/Data/Phsi_Blos/feature/%s/' % 'DIP'
# fin_model = os.path.join(f2dirout, oldfile, '_my_model.h5')
# dirout_result_test = '/home/19jjhnenu/Data/Phsi_Blos/result/5CV_1_DIP_posi/%d' % cv
# check_path(dirout_result_test)
# savepredict(fin_test, dirout_feature, fin_model, dirout_result_test, batch_size=500,onehot = onehot)
'''
testing on all.txt in Imex
'''
# fin_test = 'file/8ImexPredict/4pair.tsv'
# dirout_feature = '/home/19jjhnenu/Data/Phsi_Blos/feature/%s/' % 'IMEx'
# fin_model = os.path.join(f2dirout, oldfile, '_my_model.h5')
# dirout_result_test = '/home/19jjhnenu/Data/Phsi_Blos/result/5CV_1_IMEx_posi/%d' % cv
# check_path(dirout_result_test)
# savepredict(fin_test, dirout_feature, fin_model, dirout_result_test, batch_size=500,onehot = onehot)
'''
IMEx + -
'''
# fin_test = 'file/8ImexPredict/predict/0/all.txt'
# dirout_feature = '/home/19jjhnenu/Data/Phsi_Blos/feature/%s/' % 'IMEx'
# fin_model = os.path.join(f2dirout, oldfile, '_my_model.h5')
# dirout_result_test = '/home/19jjhnenu/Data/Phsi_Blos/result/5CV_1_IMEx/%d' % cv
# check_path(dirout_result_test)
# savepredict(fin_test, dirout_feature, fin_model, dirout_result_test, batch_size=500,onehot = onehot)
'''
testing on DIP all.txt in DIP/data/Human
'''
# for eachfile in ['Ecoli', 'Human', 'Mus', 'SC']:
# fin_test = 'file/8DIPPredict/data/%s/2pair.tsv'%eachfile
# dirout_feature = '/home/19jjhnenu/Data/Phsi_Blos/feature/%s/' % 'DIP'
# fin_model = os.path.join(f2dirout, oldfile, '_my_model.h5')
# dirout_result_test = '/home/19jjhnenu/Data/Phsi_Blos/result/5CV_1_DIP_%s/%d' % (eachfile,cv)
# check_path(dirout_result_test)
# savepredict(fin_test, dirout_feature, fin_model, dirout_result_test, batch_size=500,onehot = onehot)
'''
#
# fin_pair = 'file/8ImexPredict/data/2pair.tsv'
# fin_fasta = 'file/8ImexPredict/data/2pair.fasta'
# eachdir = 'IMEx'
# dir_feature_db = '/home/19jjhnenu/Data/Phsi_Blos/featuredb/%s/' % eachdir
# dirout_feature = '/home/19jjhnenu/Data/Phsi_Blos/feature/%s/' % eachdir
# # getFeature(fin_pair, fin_fasta, dir_feature_db, dirout_feature)
'''
train
'''
eachdir = 'benchmark'
dirout_feature = '/home/19jjhnenu/Data/Phsi_Blos/feature/%s/' % eachdir
dir_in = dirout_feature
eachdir = 'benchmark'
dirout = '/home/19jjhnenu/Data/Phsi_Blos/result/%s/' % eachdir
check_path(dirout)
validate = {}
validate['fin_pair'] = 'file/4train/0/validate.txt'
validate['dir_in'] = dir_in
onehot = False
import os
import tensorflow as tf
gpu_id = '0,1,2,3'
# gpu_id = '4,5,6,7'
os.environ['CUDA_VISIBLE_DEVICES'] = str(gpu_id)
os.system('echo $CUDA_VISIBLE_DEVICES')
tf_config = tf.compat.v1.ConfigProto()
tf_config.gpu_options.allow_growth = True
SC
uniprotPairFromImex
'''
# # mpf = '/home/jjhnenu/data/PPI/release/criteria/allcession_KW-0472_189043.list'
# tmpf = '/home/jjhnenu/data/PPI/release/criteria/allcession_KW-0812_131609.list'
# spf = '/home/jjhnenu/data/PPI/release/criteria/allcession_soluble_614454.list'
# finPair = '/home/jjhnenu/data/PPI/release/otherdata/Ecoli/2Ecoli20170205_id_pair_12246.txt'
# foutPair = '/home/jjhnenu/data/PPI/release/otherdata/Ecoli/TMP_SP/TMP_SP.txt'
# _path,_fname = os.path.split(foutPair)
# check_path(_path)
# getTmp_SpPair(tmpf,spf,finPair, foutPair,type1='TMP',type2='SP',crossover = True)
'''
count pair
'''
# dirout_pair = r'E:\githubCode\BioComputeCodeStore\JiangJiuhong\data\PPI\release\otherdata\DIP\_TMP_nonTMP_qualified_drop_positive'
# check_path(dirout_pair)
# for eachfile in os.listdir(dirout_pair):
# fout_pair = os.path.join(dirout_pair, eachfile)
# countpair(fout_pair)
'''
imex 2020 0708
'''
tmpf = '/home/jjhnenu/data/PPI/release/criteria/allcession_KW-0812_131609.list'
spf = '/home/jjhnenu/data/PPI/release/criteria/allcession_soluble_614454.list'
finPair = '/home/jjhnenu/data/PPI/release/otherdata/uniprotPiarFromImex20200709/getpair15955.csv'
foutPair = '/home/jjhnenu/data/PPI/release/otherdata/uniprotPiarFromImex20200709/TMP_SP/TMP_SP.txt'
_path,_fname = os.path.split(foutPair)
check_path(_path)
getTmp_SpPair(tmpf,spf,finPair, foutPair,type1='TMP',type2='SP',crossover = True)
def main(limit=5):
# if __name__ == '__main__':
# limit = 0
# limit = 5
onehot = False
'''
load feature
'''
print('Loading feature...')
'''
pssm
'''
# model_type = Param.CNN2D
# tpssmd = TmpPSSMData()
# # (x_train, y_train), (x_test, y_test) = tpssmd.loadPAN(inPDir, inNDir, limit=limit)
#
# fin_pair = '/home/jjhnenu/data/PPI/release/data/p_fp_fw_2_1_1/all.txt'
# dir_in = '/home/jjhnenu/data/PPI/release/feature/pssm_feature_2D/p_fp_fw_2_1_1/all/'
# fout = '/home/jjhnenu/data/PPI/release/result/pssm_feature_2D/p_fp_fw_2_1_1/all'
# check_path(fout)
# (x_train, y_train), (x_test, y_test) = tpssmd.load(fin_pair,dir_in,limit=limit)
'''
seq1D
'''
# model_type = Param.CNN1D
# fin_pair = '/home/jjhnenu/data/PPI/release/data/p_fp_fw_2_1_1/all.txt'
# dir_in = '/home/jjhnenu/data/PPI/release/feature/seq_feature_1D/p_fp_fw_2_1_1/all/'
# fout = '/home/jjhnenu/data/PPI/release/result/seq_feature_1D/p_fp_fw_2_1_1/all/'
'''
seq1D onehot
'''
# model_type = Param.CNN1D_OH
# onehot = True
# fin_pair = '/home/jjhnenu/data/PPI/release/data/p_fp_fw_2_1_1/all.txt'
# dir_in = '/home/jjhnenu/data/PPI/release/feature/seq_feature_1D/p_fp_fw_2_1_1/all'
# fout = '/home/jjhnenu/data/PPI/release/result/seq_feature_1D_onehot/p_fp_fw_2_1_1/all/'
'''
test seq2D
Could not load dynamic library 'libnvinfer.so.6'; dlerror: libnvinfer.so.6: cannot open shared object file: No such file or directory
2020-04-30 21:41:50.315008: W tensorflow/stream_executor/platform/default/dso_loader.cc:55] Could not load dynamic library 'libnvinfer_plugin.so.6'; dlerror: libnvinfer_plugin.so.6: cannot open shared object file: No such file or directory
2020-04-30 21:41:50.315029: W tensorflow/compiler/tf2tensorrt/utils/py_utils.cc:30] Cannot dlopen some TensorRT libraries. If you would like to use Nvidia GPU with TensorRT, please make sure the missing libraries mentioned above are installed properly
'''
onehot = True
model_type = Param.CNN2D
fin_pair = '/home/jjhnenu/data/PPI/release/pairdata/p_fp_fw_2_1_1/all.txt'
dir_in = '/home/jjhnenu/data/PPI/release/feature/seq_feature_2D/p_fp_fw_2_1_1/all/'
fout = '/home/jjhnenu/data/PPI/release/result/seq_feature_2D/p_fp_fw_2_1_1/all/'
'''
pssm hstack
'''
# model_type = Param.CNN2D
# fin_pair = '/home/jjhnenu/data/PPI/release/data/p_fp_fw_2_1_1/all.txt'
# dir_in = '/home/jjhnenu/data/PPI/release/feature/pssm_feature_2D_hstack/p_fp_fw_2_1_1/all/'
# fout = '/home/jjhnenu/data/PPI/release/result/pssm_feature_2D_hstack/p_fp_fw_2_1_1/all/'
'''
pssm 400
'''
# model_type = Param.CNN1D
# fin_pair = '/home/jjhnenu/data/PPI/release/data/p_fp_fw_2_1_1/all.txt'
# dir_in = '/home/jjhnenu/data/PPI/release/feature/pssm400_feature_1D/p_fp_fw_2_1_1/all/'
# fout = '/home/jjhnenu/data/PPI/release/result/pssm400_feature_1D/p_fp_fw_2_1_1/all/'
check_path(fout)
(x_train, y_train), (x_test, y_test) = BaseData().load(fin_pair,
dir_in,
limit=limit,
onehot=onehot)
print('Build and fit model...')
print('x_train.shape[1:]', x_train.shape[1:])
mm = MyModel(
model_type=model_type,
input_shape=x_train.shape[1:],
filters=250,
kernel_size=3,
pool_size=2,
hidden_dims=250,
batch_size=100,
epochs=80,
)
mm.process(fout, x_train, y_train, x_test, y_test)
print('save result to %s' % fout)
