def Protein_gen(data, Proteingroup):
import numpy as np
from pydpi.pypro import PyPro
protein = PyPro()
HP_list, D_list = [], []
for ii in range(len(data)):
p = data[ii]
protein.ReadProteinSequence(p)
keys, values = [], []
for jj in Proteingroup:
if jj == '0': #All descriptors 2049
res = protein.GetALL()
elif jj == '1': #amino acid composition 20
res = protein.GetAAComp()
elif jj == '2': #dipeptide composition 400
res = protein.GetDPComp()
elif jj == '3': #Tripeptide composition 8000
res = protein.GetTPComp()
elif jj == '4': #Moreau-Broto autocorrelation 240
res = protein.GetMoreauBrotoAuto()
elif jj == '5': #Moran autocorrelation 240
res = protein.GetMoranAuto()
elif jj == '6': #Geary autocorrelation 240
res = protein.GetGearyAuto()
elif jj == '7': #composition,transition,distribution 21+21+105
res = protein.GetCTD()
elif jj == '8': #conjoint triad features 343
res = protein.GetTriad()
elif jj == '9': #sequence order coupling number 60
res = protein.GetSOCN(30)
elif jj == '10': #quasi-sequence order descriptors 100
res = protein.GetQSO()
elif jj == '11': #pseudo amino acid composition 50
res = protein.GetPAAC(30)
keys.extend(res.keys())
values.extend(res.values())
if ii == 0:
HP_list = keys
D_list.append(values)
else:
D_list.append(values)
D_Pro = np.zeros((len(D_list), len(HP_list)), dtype=float)
for k in range(len(D_list)):
D_Pro[k, :] = D_list[k]
#Variance threshold std > 0.01
import Descriptors_Selection as DesSe
ind_var = DesSe.VarinceThreshold(D_Pro)
D_Pro = D_Pro[:, ind_var]
HP_list = np.array(HP_list)[ind_var]
H_Pro = np.reshape(HP_list, (1, len(HP_list)))
Array_Pro = np.append(H_Pro, D_Pro, axis=0)
return Array_Pro
def descriptor_generator(self, protein_sequence):
obj_PyPro = PyPro()
obj_PyPro.ReadProteinSequence(protein_sequence)
ds1 = obj_PyPro.GetAAComp()
ds2 = obj_PyPro.GetDPComp()
ds3 = obj_PyPro.GetPAAC(lamda=30)
ds4 = obj_PyPro.GetCTD()
ds5 = obj_PyPro.GetQSO()
ds6 = obj_PyPro.GetTriad()
ds_all = []
# This is use to append since sequentially add .. otherwise update() function update not sequentially
for ds in (ds1, ds2, ds3, ds4, ds5, ds6):
ds_all.append(ds)
return ds_all
def Decriptor_generator(self, ps):
protein = PyPro()
protein.ReadProteinSequence(ps)
moran = protein.GetPAAC(lamda=5,weight=0.5)
DS_1 = protein.GetAPAAC(lamda=5,weight=0.5)
DS_2 = protein.GetCTD()
DS_3 = protein.GetDPComp()
DS_4 = protein.GetGearyAuto()
DS_5 = protein.GetMoranAuto()
DS_6 = protein.GetMoreauBrotoAuto()
DS_7 = protein.GetQSO()
DS_8 = protein.GetSOCN()
DS_9 = protein.GetTPComp()
DS_ALL = {}
for DS in (DS_1,DS_2,DS_3,DS_4,DS_5,DS_6,DS_7,DS_8,DS_9,moran):
DS_ALL.update(DS)
return DS_ALL
pca = PCA(n_components=2)
X_r = pca.fit(X).transform(X)
plt.figure()
for color, i, target_name in zip(['navy', 'darkorange'], [0, 1], ['0', '1']):
plt.scatter(X_r[y == i, 0], X_r[y == i, 1], color=color, alpha=.8, lw=2,label=target_name)
plt.legend(loc='best', shadow=False, scatterpoints=1)
plt.title('PCA of hmmscan df')
# Collect physicochemical features homology -- pydpi ##
print 'Collecting physicochemical stats per protein...'
with open('./physicochem_annot_training.csv', 'w') as f:
for record in SeqIO.parse(all_fasta, "fasta"):
protein = PyPro()
protein.ReadProteinSequence(str(record.seq))
desc = protein.GetGearyAuto()
desc2 = protein.GetDPComp()
z = desc.copy()
z.update(desc2)
len_p = str(len(record.seq))
label = str(record.description).strip().split(':')[1]
id = str(record.id)
row = [id, label, len_p] + [str(i) for i in z.values()]
f.write(','.join(row) + '\n')
df_desc = pd.read_table('./physicochem_annot_training.csv', header=None, sep=',')
header = z.keys()
df_desc.columns = ['id', 'label', 'seq_length'] + header
print 'Pydpi table: {}'.format(df_desc.columns)
# plot
def Decriptor_generator(infile, lamda, weight, maxlag, destype, out_file):
list_pep_name = []
f = open(infile)
lines = f.readlines()
for line in lines:
if ">" in line:
pass
else:
list_pep_name.append(line.strip('\n'))
out_df = pd.DataFrame()
for seq in list_pep_name:
protein = PyPro()
protein.ReadProteinSequence(seq)
if destype == "GetAAComp":
DS = protein.GetAAComp()
df = pd.DataFrame(DS, index=[0])
elif destype == "GetDPComp":
DS = protein.GetDPComp()
df = pd.DataFrame(DS, index=[0])
elif destype == "GetTPComp":
DS = protein.GetTPComp()
df = pd.DataFrame(DS, index=[0])
elif destype == "GetMoreauBrotoAuto":
DS = protein.GetMoreauBrotoAuto()
df = pd.DataFrame(DS, index=[0])
elif destype == "GetMoranAuto":
DS = protein.GetMoranAuto()
df = pd.DataFrame(DS, index=[0])
elif destype == "GetGearyAuto":
DS = protein.GetGearyAuto()
df = pd.DataFrame(DS, index=[0])
elif destype == "GetCTD":
DS = protein.GetCTD()
df = pd.DataFrame(DS, index=[0])
elif destype == "GetPAAC":
DS = protein.GetPAAC(lamda=int(lamda), weight=float(weight))
df = pd.DataFrame(DS, index=[0])
elif destype == "GetAPAAC":
DS = protein.GetAPAAC(lamda=int(lamda), weight=float(weight))
df = pd.DataFrame(DS, index=[0])
elif destype == "GetSOCN":
DS = protein.GetSOCN(maxlag=int(maxlag))
df = pd.DataFrame(DS, index=[0])
elif destype == "GetQSO":
DS = protein.GetQSO(maxlag=int(maxlag), weight=float(weight))
df = pd.DataFrame(DS, index=[0])
elif destype == "GetTriad":
DS = protein.GetTriad()
df = pd.DataFrame(DS, index=[0])
elif destype == "All":
DS1 = protein.GetAAComp()
DS2 = protein.GetDPComp()
DS3 = protein.GetTPComp()
DS4 = protein.GetMoreauBrotoAuto()
DS5 = protein.GetMoranAuto()
DS6 = protein.GetGearyAuto()
DS7 = protein.GetCTD()
DS8 = protein.GetPAAC(lamda=int(lamda), weight=float(weight))
DS9 = protein.GetAPAAC(lamda=int(lamda), weight=float(weight))
DS10 = protein.GetSOCN(maxlag=int(maxlag))
DS11 = protein.GetQSO(maxlag=int(maxlag), weight=float(weight))
DS12 = protein.GetTriad()
DS = {}
for D in (DS1, DS2, DS3, DS4, DS5, DS6, DS7, DS8, DS9, DS10, DS11,
DS12):
print(D)
DS.update(D)
df = pd.DataFrame(DS, index=[0])
if destype == 'BinaryDescriptor':
out_df = BinaryDescriptor(list_pep_name)
else:
out_df = pd.concat([out_df, df], axis=0)
out_df.to_csv(out_file, index=False, sep='\t')