A question about the PSTNPss function in util/FileProcessing.py

[DellCode233]

Hello, I am interested in your project, but I have some questions when reading your source code. I hope you can help me answer them.

My question is about the PSTNPss function in util/FileProcessing.py.
Question 1: I see that in this function, you subtract one from the total number of samples for the corresponding label and subtract one from the trinucleotide count at the corresponding location. I don’t understand the purpose and principle of doing this.

p_num, n_num = positive_number, negative_number
po_number = matrix_po[j][order[sequence[j: j + 3]]]
if i[0] in positive_key and po_number > 0:
    po_number -= 1
    p_num -= 1
ne_number = matrix_ne[j][order[sequence[j: j + 3]]]
if i[0] in negative_key and ne_number > 0:
    ne_number -= 1
    n_num -= 1

Question 2: Secondly, this function uses different processing methods for the training dataset and the testing dataset. In the training dataset, you perform the above subtraction operation, but not in the testing dataset. I don’t understand why there is such a difference. I have attached your code snippet for your convenience. Thank you for your time and help!

    def PSTNPss(self):
        try:
            if not self.is_equal:
                self.error_msg = 'PSTNPss descriptor need fasta sequence with equal length.'
                return False

            fastas = []
            for item in self.fasta_list:
                if item[3] == 'training':
                    fastas.append(item)
                    fastas.append([item[0], item[1], item[2], 'testing'])
                else:
                    fastas.append(item)

            for i in fastas:
                if re.search('[^ACGT-]', i[1]):
                    self.error_msg = 'Illegal character included in the fasta sequences, only the "ACGT[U]" are allowed by this encoding scheme.'
                    return False

            encodings = []
            header = ['SampleName', 'label']
            for pos in range(len(fastas[0][1]) - 2):
                header.append('Pos.%d' % (pos + 1))
            encodings.append(header)

            positive = []
            negative = []
            positive_key = []
            negative_key = []
            for i in fastas:
                if i[3] == 'training':
                    if i[2] == '1':
                        positive.append(i[1])
                        positive_key.append(i[0])
                    else:
                        negative.append(i[1])
                        negative_key.append(i[0])

            nucleotides = ['A', 'C', 'G', 'T']
            trinucleotides = [n1 + n2 + n3 for n1 in nucleotides for n2 in nucleotides for n3 in nucleotides]
            order = {}
            for i in range(len(trinucleotides)):
                order[trinucleotides[i]] = i

            matrix_po = self.CalculateMatrix(positive, order)
            matrix_ne = self.CalculateMatrix(negative, order)

            positive_number = len(positive)
            negative_number = len(negative)

            for i in fastas:
                if i[3] == 'testing':
                    name, sequence, label = i[0], i[1], i[2]
                    code = [name, label]
                    for j in range(len(sequence) - 2):
                        if re.search('-', sequence[j: j + 3]):
                            code.append(0)
                        else:
                            p_num, n_num = positive_number, negative_number
                            po_number = matrix_po[j][order[sequence[j: j + 3]]]
                            if i[0] in positive_key and po_number > 0:
                                po_number -= 1
                                p_num -= 1
                            ne_number = matrix_ne[j][order[sequence[j: j + 3]]]
                            if i[0] in negative_key and ne_number > 0:
                                ne_number -= 1
                                n_num -= 1
                            code.append(po_number / p_num - ne_number / n_num)
                            # print(sequence[j: j+3], order[sequence[j: j+3]], po_number, p_num, ne_number, n_num)
                    encodings.append(code)
            self.encoding_array = np.array([])
            self.encoding_array = np.array(encodings, dtype=str)
            self.column = self.encoding_array.shape[1]
            self.row = self.encoding_array.shape[0] - 1
            del encodings
            if self.encoding_array.shape[0] > 1:
                return True
            else:
                return False
        except Exception as e:
            self.error_msg = str(e)
            return False