def quantify_data():
sets = all_set_filepath()
classification_to_number = {"H":0, "E": 1, "C":2}
pc = get_given_principal_components()
for dataset in sets:
train_classification_file = []
train_sequence_file = os.path.join(dataset,"SVM_Training_Categorical_Feature.txt")
train_sequence = get_string(train_sequence_file)
for i in range(5):
filename = "SVM_Training_Classification_Transformed_Rule_" + str(i+1) + ".txt"
train_classification_file.append(os.path.join(dataset,filename))
for file in train_classification_file:
print file
number = file[-5]
tf = get_tendency_factor(train_sequence_file, file)
classification = get_string(file)
filename = os.path.join(dataset,"SVM_Training_Feature_" + number + ".txt")
create_file(filename)
filename = os.path.join(dataset,"SVM_Training_Label_" + number + ".txt")
create_file(filename)
list = []
for i in range(len(classification)):
code = tf[train_sequence[i]] + pc[train_sequence[i]]
a,b,c,d,e,f = code
code = (a,b,c,d,e,f,classification_to_number[classification[i]])
list.append(code)
for a,b,c,d,e,f,g in list:
file = open(os.path.join(dataset,"SVM_Training_Feature_" + number + ".txt"), "a+")
file.write(str(round(a,2))+","+str(round(b,2))+","+str(round(c,2))+","+str(round(d,2))+","+str(round(e,2))+","+str(round(f,2))+"\n")
file.close()
file = open(os.path.join(dataset,"SVM_Training_Label_" + number + ".txt"), "a+")
file.write(str(g) + "\n")
file.close()
def predict_and_test(): #better named results
create_file("Overall_Results.txt")
sets = all_set_filepath()
for i in range(len(sets)):
similarity_prediction(sets[i], 15, 0.9)
for dataset in sets:
svm_prediction(dataset)
get_accuracy_state(dataset)
def svm_prediction(setpath):
number = setpath[-1]
pc = get_given_principal_components()
test_sequence_file = os.path.join(setpath,
"SVM_Testing_Categorical_Feature.txt")
with open(test_sequence_file) as f:
test_proteins = {
line.split("|")[0]: line.split("|")[1][:-1]
for line in f
}
for m in range(5):
filename = "SVM_Set_" + number + "_" + str(m + 1) + ".pkl"
print filename
classifier = joblib.load(filename)
print classifier
test_classification_file = os.path.join(
setpath, "SVM_Testing_Classification_Transformed_Rule_" +
str(m + 1) + ".txt")
tf = get_tendency_factor(test_sequence_file, test_classification_file)
filename = "Prediction_Based_On_Similarity_" + str(m + 1) + ".txt"
print filename
print "started", strftime("%Y-%m-%d %H:%M:%S", gmtime())
with open(os.path.join(setpath, filename)) as f:
predictions = {
line.split("|")[0]: line.split("|")[1][:-1]
for line in f
}
print len(predictions)
create_file(os.path.join(setpath, "temp.txt"))
for name, item in predictions.iteritems():
for i in range(len(item)):
pass
if item[i] == " ":
code = tf[test_proteins[name][i]] + pc[test_proteins[name]
[i]]
a, b, c, d, e, f = code
code = (round(a, 2), round(b, 2), round(c, 2), round(d, 2),
round(e, 2), round(f, 2))
nparray = np.array(list(code))
reshaped_array = nparray.reshape(1, -1)
svm_output = int(
classifier.predict(reshaped_array).tolist()[0])
svm_output_to_categorical = {0: "H", 1: "E", 2: "C"}
item = item[:i] + svm_output_to_categorical[
svm_output] + item[i + 1:]
predictions[name] = item
file = open(os.path.join(setpath, "temp.txt"), "a+")
for name, item in predictions.iteritems():
file.write(name + "|" + item + "\n")
file.close()
os.remove(os.path.join(setpath, filename))
os.rename(os.path.join(setpath, "temp.txt"),
os.path.join(setpath, filename))
print "completed", strftime("%Y-%m-%d %H:%M:%S", gmtime())
def purify_data():
create_file("new_sequence.txt")
create_file("new_secondary_structure.txt")
sequence = get_all_sequence()
new_seq = remove_unavailable(sequence)
write_file(new_seq)
update_secondary_structure(new_seq.keys())
os.remove("Training_Sequence.txt")
os.remove("Training_Classification.txt")
os.rename("new_sequence.txt", "Training_Sequence.txt")
os.rename("new_secondary_structure.txt", "Training_Classification.txt")
def split_categorical_data(proportion):
path = os.path.realpath("Training_Set_1")
while os.path.exists(path):
path = path[:-1] + str(int(path[-1])+1)
os.mkdir(path)
sequence = get_all_sequence()
classification = get_all_classification()
if len(classification.keys()) != len(sequence.keys()):
print "classification and sequence are of different length"
return
if proportion*len(classification) < 1 and proportion*len(classification) > 0:
print "proportion too small"
return
elif proportion >= 1 or proportion <= 0:
print "proportion must be within (0,1)"
return
create_file(os.path.join(path,"SVM_Training_Categorical_Feature.txt"))
create_file(os.path.join(path,"SVM_Training_Categorical_Label.txt"))
create_file(os.path.join(path,"SVM_Testing_Categorical_Feature.txt"))
create_file(os.path.join(path,"SVM_Testing_Categorical_Label.txt"))
create_file(os.path.join(path,"SVM_Reference_Categorical_Feature.txt"))
create_file(os.path.join(path,"SVM_Reference_Categorical_Label.txt"))
keylist = classification.keys()
shuffle(keylist)
training_length = proportion*len(classification)
for i in range(len(keylist)):
if i < training_length:
file = open(os.path.join(path,"SVM_Training_Categorical_Feature.txt"),"a+")
file.write(keylist[i]+"|"+sequence[keylist[i]]+"\n")
file.close()
file = open(os.path.join(path,"SVM_Training_Categorical_Label.txt"),"a+")
file.write(keylist[i]+"|"+classification[keylist[i]] + "\n")
file.close()
elif i < 2*training_length:
file = open(os.path.join(path,"SVM_Reference_Categorical_Feature.txt"),"a+")
file.write(keylist[i]+"|"+sequence[keylist[i]]+"\n")
file.close()
file = open(os.path.join(path,"SVM_Reference_Categorical_Label.txt"),"a+")
file.write(keylist[i]+"|"+classification[keylist[i]] + "\n")
file.close()
else:
file = open(os.path.join(path,"SVM_Testing_Categorical_Feature.txt"),"a+")
file.write(keylist[i]+"|"+sequence[keylist[i]]+"\n")
file.close()
file = open(os.path.join(path,"SVM_Testing_Categorical_Label.txt"),"a+")
file.write(keylist[i]+"|"+classification[keylist[i]] + "\n")
file.close()
if length_of(os.path.join(path,"SVM_Training_Categorical_Feature.txt")) + length_of(os.path.join(path,"SVM_Testing_Categorical_Feature.txt")) + length_of(os.path.join(path,"SVM_Reference_Categorical_Feature.txt"))== len(keylist):
print "successful separation"
def similarity_prediction(setpath, window_size, threshold):
file = open("Overall_Results.txt", "a+")
file.write("Trial " + setpath[-1] + ":\n")
file.close()
if window_size % 2 == 0:
sidelength = window_size / 2
else:
sidelength = (window_size - 1) / 2
for m in range(5):
with open(
os.path.join(setpath,
"SVM_Reference_Categorical_Feature.txt")) as f:
list = [line.split("|")[0] for line in f]
filename = "SVM_Reference_Classification_Transformed_Rule_" + str(
m + 1) + ".txt"
with open(os.path.join(setpath, filename)) as f:
list2 = [line.split("|")[0] for line in f]
if list != list2:
return
with open(
os.path.join(setpath,
"SVM_Reference_Categorical_Feature.txt")) as f:
refs = {line.split("|")[0]: line.split("|")[1][:-1] for line in f}
with open(os.path.join(setpath,
"SVM_Testing_Categorical_Feature.txt")) as f:
test = {line.split("|")[0]: line.split("|")[1][:-1] for line in f}
filename = "SVM_Reference_Classification_Transformed_Rule_" + str(
m + 1) + ".txt"
with open(os.path.join(setpath, filename)) as f:
ref_structure = {
line.split("|")[0]: line.split("|")[1][:-1]
for line in f
}
unpredictable_amino_acids = 0
total_amino_acids = 0
bad_proteins = {}
predicted_proteins = {}
test_length = 0
for protein in test.itervalues():
test_length += len(protein)
print test_length
for name, protein in test.iteritems():
prediction = ""
for i in range(len(protein)):
total_amino_acids += 1
if i - sidelength < 0 or i + sidelength >= len(protein):
prediction += " "
unpredictable_amino_acids += 1
continue
protein_sequence = protein[i - sidelength:i + sidelength + 1]
protein_sequence_3d = Complicated_Math.get_coordinate(
(1, 2, 7), protein_sequence)
similar = similar_amino_acid(protein_sequence_3d, refs,
window_size, threshold)
if similar == None:
prediction += " "
unpredictable_amino_acids += 1
continue
else:
sim_name, index = similar
prediction += ref_structure[sim_name][index]
if len(prediction) != len(protein):
print "Protein", name, "predicted secondary structure length is problematic."
bad_proteins[name] = prediction
continue
predicted_proteins[name] = prediction
print str(total_amino_acids), "/", str(test_length)
file = open("Overall_Results.txt", "a+")
file.write(
"When using translation method " + str(m + 1) + ", " +
str(round(unpredictable_amino_acids /
float(total_amino_acids), 1)) +
"% of Proteins cannot be predicted via similarity analysis.\n")
file.close()
filename = "Prediction_Based_On_Similarity_" + str(m + 1) + ".txt"
create_file(os.path.join(setpath, filename))
file = open(os.path.join(setpath, filename), "a+")
for name, item in predicted_proteins.iteritems():
file.write(name + "|" + item + "\n")
file.close()
filename = "Prediction_Pure_SVM_" + str(m + 1) + ".txt"
create_file(os.path.join(setpath, filename))
file = open(os.path.join(setpath, filename), "a+")
for name, item in predicted_proteins.iteritems():
file.write(name + "|")
for i in item:
file.write(" ")
file.write("\n")
file.close()