def make_NN(sw, dataset, groups, without):
ann = create_ann(2, sw, len(codes.aac))
inputs_train = []
outputs_train = []
protCodes = []
for i in range(len(groups)):
if i == without:
continue
else:
for s in groups[i]:
protCodes.append(s)
for p in protCodes:
sec = dataset[p]['sec']
prim = dataset[p]['prim']
#prim = prim[-2:]
primx = inOutFunctions.merge_sequences(prim)
ins, outs = inOutFunctions.convert_inputNN(sec, primx, sw)
for q in ins:
inputs_train.append(q)
for q in outs:
outputs_train.append(q)
ann = train_ann(ann, inputs_train, outputs_train)
return ann
def make_NN2(ann, sw, dataset, groups, without):
ann2 = create_ann(2, sw, 3)
inputs_train = []
outputs_train = []
protCodes = []
for s in groups[without]:
protCodes.append(s)
for p in protCodes:
sec = dataset[p]['sec']
prim = dataset[p]['prim']
#prim = prim[-2:]
prim = inOutFunctions.merge_sequences(prim)
ins, outs = inOutFunctions.convert_inputNN(sec, prim, sw)
pred = ann.predict(np.array(ins, np.float32))
ins, outs = inOutFunctions.convert_inputNN2(sec, pred, sw)
for q in ins:
inputs_train.append(q)
for q in outs:
outputs_train.append(q)
ann2 = train_ann(ann2, inputs_train, outputs_train)
return ann2
def make_SVM_1(sw, dssp, x, dataset):
inputs_train = []
outputs_train = []
#read dataset file
f = open(dataset, 'r')
for i in range(x):
desc = f.readline().strip()
primlen = int(desc.split('#')[1])
prim = []
for j in range(primlen):
prim.append(f.readline().strip())
sec = f.readline().strip()
#prim = prim[-2:]
primx = inOutFunctions.merge_sequences(prim)
ins, outs = inOutFunctions.prepare_input_forX(sec, primx, sw, dssp)
for q in ins:
inputs_train.append(q)
for q in outs:
outputs_train.append(q)
f.close()
clf = svm.SVC(C=2.5, gamma=0.05)
clf.fit(inputs_train, outputs_train)
return clf
def test_NN(ann, sw, dataset, groups, without):
sum = 0
q3 = 0
qh = 0
qhp = 0
qe = 0
qep = 0
qc = 0
qcp = 0
sovh, zh = 0, 0.01
sove, ze = 0, 0.01
sovc, zc = 0, 0.01
z = 126/7
protCodes = []
for s in groups[without]:
protCodes.append(s)
for p in protCodes:
sec = dataset[p]['sec']
prim = dataset[p]['prim']
#prim = prim[-2:]
prim = inOutFunctions.merge_sequences(prim)
ins, outs = inOutFunctions.convert_inputNN(sec, prim, sw)
pred = ann.predict(np.array(ins, np.float32))
predx = []
for p in pred:
predx.append(winner(p))
pred = inOutFunctions.display_result(predx, codes.alphabeth)
sum += measurePrediction.compare(pred, sec)
q3 += measurePrediction.calcQ3(pred, sec)
qh += measurePrediction.calcQ(pred, sec, 'H')
qhp += measurePrediction.calcQpred(pred, sec, 'H')
qe += measurePrediction.calcQ(pred, sec, 'E')
qep += measurePrediction.calcQpred(pred, sec, 'E')
qc += measurePrediction.calcQ(pred, sec, 'C')
qcp += measurePrediction.calcQpred(pred, sec, 'C')
_sovh = measurePrediction.calcSOV(pred, sec, 'H')
_sove = measurePrediction.calcSOV(pred, sec, 'E')
_sovc = measurePrediction.calcSOV(pred, sec, 'C')
if _sovh != None:
sovh += _sovh
zh += 1
if _sove != None:
sove += _sove
ze += 1
if _sovc != None:
sovc += _sovc
zc += 1
return (q3/z, qh/z, qhp/z, qe/z, qep/z, qc/z, qcp/z, sovh/zh, sove/ze, sovc/zc)
def test_SMV_1(sw, dssp, w, clfx, z):
f = open('rs126.fa', 'r')
cq = 0
cqp = 0
for i in range(z+w):
desc = f.readline().strip()
primlen = int(desc.split('#')[1])
prim = []
for j in range(primlen):
prim.append(f.readline().strip())
sec = f.readline().strip()
#prim = prim[-2:]
if i >= z:
primx = inOutFunctions.merge_sequences(prim)
ins, outs = inOutFunctions.prepare_input_forX(sec, primx, sw, dssp)
pred = inOutFunctions.display_result(clfx.predict(np.array(ins, np.float32)), {0:'X', 1:dssp})
print sec
print pred
print "\n"
cq += measurePrediction.calcQ(pred, sec, dssp)
cqp += measurePrediction.calcQpred(pred, sec, dssp)
return (cq/w, cqp/w)
def make_SVM_3X(sw, dataset, groups, without):
inputs_train = []
outputs_train = []
protCodes = []
for i in range(len(groups)):
if i == without:
continue
else:
for s in groups[i]:
protCodes.append(s)
for p in protCodes:
sec = dataset[p]['sec']
prim = dataset[p]['prim']
prim = prim[-2:]
primx = inOutFunctions.merge_sequences(prim)
ins, outs = inOutFunctions.convert_inputX(sec, primx, sw)
for q in ins:
inputs_train.append(q)
for q in outs:
outputs_train.append(q)
clf = svm.SVC(C=1.5, gamma=0.1)
clf.fit(inputs_train, outputs_train)
return clf
def test_SVM_3(clf, z, x, sw, filename):
f = open(filename, 'r')
q3 = 0
qh = 0
qhp = 0
qe = 0
qep = 0
qc = 0
qcp = 0
sovh, zh = 0, 0.01
sove, ze = 0, 0.01
sovc, zc = 0, 0.01
for i in range(x+z):
desc = f.readline().strip()
primlen = int(desc.split('#')[1])
prim = []
for j in range(primlen):
prim.append(f.readline().strip())
sec = f.readline().strip()
#prim = prim[-2:]
if i >= x:
primx = inOutFunctions.merge_sequences(prim)
ins, outs = inOutFunctions.convert_inputX(sec, primx, sw)
pred = inOutFunctions.display_result(clf.predict(np.array(ins, np.float32)), codes.alphabeth)
q3 += measurePrediction.calcQ3(pred, sec)
qh += measurePrediction.calcQ(pred, sec, 'H')
qhp += measurePrediction.calcQpred(pred, sec, 'H')
qe += measurePrediction.calcQ(pred, sec, 'E')
qep += measurePrediction.calcQpred(pred, sec, 'E')
qc += measurePrediction.calcQ(pred, sec, 'C')
qcp += measurePrediction.calcQpred(pred, sec, 'C')
_sovh = measurePrediction.calcSOV(pred, sec, 'H')
_sove = measurePrediction.calcSOV(pred, sec, 'E')
_sovc = measurePrediction.calcSOV(pred, sec, 'C')
#print sec
#print pred
if _sovh != None:
sovh += _sovh
zh += 1
if _sove != None:
sove += _sove
ze += 1
if _sovc != None:
sovc += _sovc
zc += 1
f.close()
return (q3/z, qh/z, qhp/z, qe/z, qep/z, qc/z, qcp/z, sovh/zh, sove/ze, sovc/zc)
def test_SVM_3X(clf, sw, dataset, groups, without):
q3 = 0
qh = 0
qhp = 0
qe = 0
qep = 0
qc = 0
qcp = 0
sovh, zh = 0, 0.01
sove, ze = 0, 0.01
sovc, zc = 0, 0.01
protCodes = []
for s in groups[without]:
protCodes.append(s)
for p in protCodes:
sec = dataset[p]['sec']
prim = dataset[p]['prim']
#prim = prim[-2:]
primx = inOutFunctions.merge_sequences(prim)
ins, outs = inOutFunctions.convert_inputX(sec, primx, sw)
pred = inOutFunctions.display_result(clf.predict(np.array(ins, np.float32)), codes.alphabeth)
q3 += measurePrediction.calcQ3(pred, sec)
qh += measurePrediction.calcQ(pred, sec, 'H')
qhp += measurePrediction.calcQpred(pred, sec, 'H')
qe += measurePrediction.calcQ(pred, sec, 'E')
qep += measurePrediction.calcQpred(pred, sec, 'E')
qc += measurePrediction.calcQ(pred, sec, 'C')
qcp += measurePrediction.calcQpred(pred, sec, 'C')
_sovh = measurePrediction.calcSOV(pred, sec, 'H')
_sove = measurePrediction.calcSOV(pred, sec, 'E')
_sovc = measurePrediction.calcSOV(pred, sec, 'C')
if _sovh != None:
sovh += _sovh
zh += 1
if _sove != None:
sove += _sove
ze += 1
if _sovc != None:
sovc += _sovc
zc += 1
z = len(groups[without])
return (q3/z, qh/z, qhp/z, qe/z, qep/z, qc/z, qcp/z, sovh/zh, sove/ze, sovc/zc)
def test_SMV_2(clf, sw, dataset, groups, without, dssp):
cq = 0
cqp = 0
cc = 0
sov = 0
q = 18
qq = 0
protCodes = []
for s in groups[without]:
protCodes.append(s)
for p in protCodes:
sec = dataset[p]['sec']
prim = dataset[p]['prim']
#prim = prim[-2:]
primx = inOutFunctions.merge_sequences(prim)
ins, outs = inOutFunctions.prepare_input_forX(sec, primx, sw, dssp)
pred = inOutFunctions.display_result(clf.predict(np.array(ins, np.float32)), {0:'X', 1:dssp})
# print sec
# print pred
# print "\n"
cq += measurePrediction.calcQ(pred, sec, dssp)
cqp += measurePrediction.calcQpred(pred, sec, dssp)
cc += measurePrediction.calcC(pred, sec, dssp)
sovx = measurePrediction.calcSOV(pred, sec, dssp)
if sovx != None:
sov += sovx
qq += 1
return (cq/q, cqp/q, cc/q, sov/qq)
