Add application file

app.py

@@ -0,0 +1,290 @@
+# -*- coding: utf-8 -*-
+# @Time    : 2021/8/17 23:08
+# @Author  : Cheng Ge
+from tensorflow.keras.models import Model
+import gradio as gr
+import warnings
+warnings.filterwarnings('ignore')
+import numpy as np
+from numpy import linalg as la
+import tensorflow as tf
+from tensorflow import keras
+from tensorflow.keras import layers
+def pad_to_length(input_data: list, pad_token, max_length: int) -> list:
+    assert len(input_data) <= max_length
+    result = input_data[:]
+    for i in range(max_length - len(result)):
+        result.append(pad_token)
+    return result
+def TransDict_from_list(groups):
+    transDict = dict()
+    tar_list = ['0', '1', '2', '3', '4', '5', '6']
+    result = {}
+    index = 0
+    for group in groups:
+        g_members = sorted(group)  # Alphabetically sorted list
+        for c in g_members:
+            # print('c' + str(c))
+            # print('g_members[0]' + str(g_members[0]))
+            result[c] = str(tar_list[index])  # K:V map, use group's first letter as represent.
+        index = index + 1
+    return result
+def get_3_protein_trids():
+    nucle_com = []
+    chars = ['0', '1', '2', '3', '4', '5', '6']
+    base = len(chars)
+    end = len(chars) ** 3
+    for i in range(0, end):
+        n = i
+        ch0 = chars[n % base]
+        n = n / base
+        ch1 = chars[int(n % base)]
+        n = n / base
+        ch2 = chars[int(n % base)]
+        nucle_com.append(ch0 + ch1 + ch2)
+    return nucle_com
+def translate_sequence(seq, TranslationDict):
+    '''
+    Given (seq) - a string/sequence to translate,
+    Translates into a reduced alphabet, using a translation dict provided
+    by the TransDict_from_list() method.
+    Returns the string/sequence in the new, reduced alphabet.
+    Remember - in Python string are immutable..
+
+    '''
+    import string
+    from_list = []
+    to_list = []
+    for k, v in TranslationDict.items():
+        from_list.append(k)
+        to_list.append(v)
+    # TRANS_seq = seq.translate(str.maketrans(zip(from_list,to_list)))
+    TRANS_seq = seq.translate(str.maketrans(str(from_list), str(to_list)))
+    # TRANS_seq = maketrans( TranslationDict, seq)
+    return TRANS_seq
+def get_4_nucleotide_composition(tris, seq, pythoncount=True):
+    seq_len = len(seq)
+    tri_feature = [0] * len(tris)
+    k = len(tris[0])
+    note_feature = [[0 for cols in range(len(seq) - k + 1)] for rows in range(len(tris))]
+    if pythoncount:
+        for val in tris:
+            num = seq.count(val)
+            tri_feature.append(float(num) / seq_len)
+    else:
+        # tmp_fea = [0] * len(tris)
+        for x in range(len(seq) + 1 - k):
+            kmer = seq[x:x + k]
+            if kmer in tris:
+                ind = tris.index(kmer)
+                # tmp_fea[ind] = tmp_fea[ind] + 1
+                note_feature[ind][x] = note_feature[ind][x] + 1
+        # tri_feature = [float(val)/seq_len for val in tmp_fea]    #tri_feature type:list len:256
+        u, s, v = la.svd(note_feature)
+        for i in range(len(s)):
+            tri_feature = tri_feature + u[i] * s[i] / seq_len
+        # print tri_feature
+        # pdb.set_trace()
+
+    return tri_feature
+def BPF(seq_temp):
+    sequences = seq_temp
+    Seq1 = []
+    for i in range(len(sequences)):
+        kmer = sequences[i]
+        src_vocab = {'A': 1, 'C': 2, 'D': 3, 'E': 4, 'F': 5, 'G': 6, 'H': 7, 'I': 8, 'K': 9, 'L': 10, 'M': 11, 'N': 12, 'P': 13, 'Q': 14, 'R': 15, 'S': 16, 'T': 17, 'V': 18, 'W': 19, 'Y': 20, 'O': 21, 'U': 22, 'Z': 23, 'X': 23}
+        seq = src_vocab[kmer]
+        Seq1.append(seq)
+        seq = pad_to_length(Seq1, 0, 7)
+    fea = []
+    tem_vec = []
+    for i in range(len(seq)):
+        if seq[i] == 1:
+            tem_vec = [1,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0]
+        elif seq[i]==2:
+            tem_vec = [0,1,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0]
+        elif seq[i]==3:
+            tem_vec = [0,0,1,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0]
+        elif seq[i]==4:
+            tem_vec = [0,0,0,1,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0]
+        elif seq[i]==5:
+            tem_vec = [0,0,0,0,1,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0]
+        elif seq[i]==6:
+            tem_vec = [0,0,0,0,0,1,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0]
+        elif seq[i]==7:
+            tem_vec = [0,0,0,0,0,0,1,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0]
+        elif seq[i]==8:
+            tem_vec = [0,0,0,0,0,0,0,1,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0]
+        elif seq[i]==9:
+            tem_vec = [0,0,0,0,0,0,0,0,1,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0]
+        elif seq[i]==10:
+            tem_vec = [0,0,0,0,0,0,0,0,0,1,0,0,0,0,0,0,0,0,0,0,0,0,0,0]
+        elif seq[i]==11:
+            tem_vec = [0,0,0,0,0,0,0,0,0,0,1,0,0,0,0,0,0,0,0,0,0,0,0,0]
+        elif seq[i]==12:
+            tem_vec = [0,0,0,0,0,0,0,0,0,0,0,1,0,0,0,0,0,0,0,0,0,0,0,0]
+        elif seq[i]==13:
+            tem_vec = [0,0,0,0,0,0,0,0,0,0,0,0,1,0,0,0,0,0,0,0,0,0,0,0]
+        elif seq[i]==14:
+            tem_vec = [0,0,0,0,0,0,0,0,0,0,0,0,0,1,0,0,0,0,0,0,0,0,0,0]
+        elif seq[i]==15:
+            tem_vec = [0,0,0,0,0,0,0,0,0,0,0,0,0,0,1,0,0,0,0,0,0,0,0,0]
+        elif seq[i]==16:
+            tem_vec = [0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,1,0,0,0,0,0,0,0,0]
+        elif seq[i]==17:
+            tem_vec = [0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,1,0,0,0,0,0,0,0]
+        elif seq[i]==18:
+            tem_vec = [0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,1,0,0,0,0,0,0]
+        elif seq[i]==19:
+            tem_vec = [0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,1,0,0,0,0,0]
+        elif seq[i]==20:
+            tem_vec = [0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,1,0,0,0,0]
+        elif seq[i]==21:
+            tem_vec = [0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,1,0,0,0]
+        elif seq[i]==22:
+            tem_vec = [0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,1,0,0]
+        elif seq[i]==23:
+            tem_vec = [0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,1,0]
+        elif seq[i]==24:
+            tem_vec = [0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,1]
+        elif seq[i]==0:
+            tem_vec = [0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0]
+        fea = fea + tem_vec
+    return fea
+def transfer_label_from_prob(proba):
+    label = [1 if val >= 0.5 else 0 for val in proba]
+    return label
+def prepare_feature(file):
+    files = file.name
+    protein_seq_dict = {}
+    protein_index = 1
+    with open(files, 'r') as fp:
+        for line in fp:
+            seq = line[:-1]
+            protein_seq_dict[protein_index] = seq
+            protein_index = protein_index + 1
+
+    groups = ['AGV', 'ILFPO', 'YMTS', 'HNQW', 'RK', 'DEZ', 'CU']
+    group_dict = TransDict_from_list(groups)
+    protein_tris = get_3_protein_trids()
+    bpf = []
+    kmer = []
+    sequence = []
+
+    for i in protein_seq_dict:
+        protein_seq = translate_sequence(protein_seq_dict[i], group_dict)
+        if len(protein_seq_dict[i]) > 7:
+            aaa = protein_seq_dict[i][0:7]
+            bpf_feature = BPF(aaa)
+        else:
+            bpf_feature = BPF(protein_seq_dict[i])
+        protein_tri_fea = get_4_nucleotide_composition(protein_tris, protein_seq, pythoncount =False)
+
+        bpf.append(bpf_feature)
+        kmer.append(protein_tri_fea)
+        sequence.append(protein_seq_dict[i])
+    return np.array(bpf), np.array(kmer), np.array(sequence)
+
+class TransformerBlock(layers.Layer):
+    def __init__(self, embed_dim, num_heads, ff_dim, rate=0.1):
+        super(TransformerBlock, self).__init__()
+        self.att = layers.MultiHeadAttention(num_heads=num_heads, key_dim=embed_dim)
+        self.ffn = keras.Sequential(
+            [layers.Dense(ff_dim, activation="relu"), layers.Dense(embed_dim), ]
+        )
+        self.layernorm1 = layers.LayerNormalization(epsilon=1e-6)
+        self.layernorm2 = layers.LayerNormalization(epsilon=1e-6)
+        self.dropout1 = layers.Dropout(rate)
+        self.dropout2 = layers.Dropout(rate)
+
+    def call(self, inputs, training):
+        attn_output = self.att(inputs, inputs)
+        attn_output = self.dropout1(attn_output, training=training)
+        out1 = self.layernorm1(inputs + attn_output)
+        ffn_output = self.ffn(out1)
+        ffn_output = self.dropout2(ffn_output, training=training)
+        return self.layernorm2(out1 + ffn_output)
+class TokenAndPositionEmbedding(layers.Layer):
+    def __init__(self, maxlen, vocab_size, embed_dim):
+        super(TokenAndPositionEmbedding, self).__init__()
+        self.token_emb = layers.Embedding(input_dim=vocab_size, output_dim=embed_dim)
+        self.pos_emb = layers.Embedding(input_dim=maxlen, output_dim=embed_dim)
+
+    def call(self, x):
+        maxlen = tf.shape(x)[-1]
+        positions = tf.range(start=0, limit=maxlen, delta=1)
+        positions = self.pos_emb(positions)
+        x = self.token_emb(x)
+        return x + positions
+
+def ACP_DL(file):
+    data_dim = 511
+    timesteps = 1
+    len_seq_max = 18
+    bpf, kmer, sequence = prepare_feature(file)
+    Seq2 = []
+
+    for m in sequence:
+        Seq1 = []
+        for i in range(len(m)):
+            subq = m[i]
+            src_vocab = {'A': 1, 'C': 2, 'D': 3, 'E': 4, 'F': 5, 'G': 6, 'H': 7, 'I': 8, 'K': 9, 'L': 10, 'M': 11, 'N': 12, 'P': 13, 'Q': 14, 'R': 15, 'S': 16, 'T': 17, 'V': 18, 'W': 19, 'Y': 20,'O': 21, 'U': 22, 'Z': 23, 'X': 24 }
+            seq = src_vocab[subq]
+            Seq1.append(seq)
+
+            if len(Seq1) > len_seq_max:
+                Seq1 = Seq1[0:len_seq_max]
+            else:
+                Seq1 = Seq1
+
+            seq = pad_to_length(Seq1, 0, len_seq_max)
+        Seq2.append(seq)
+    Seq2 = np.array(Seq2)
+    X = np.concatenate((bpf, kmer), axis=1)
+    X = np.reshape(X, (len(X), timesteps, data_dim))
+    all_prob = {}
+    all_prob[0] = []
+    test1 = np.array(X)
+    test2 = np.array(Seq2)
+
+    Transformer_input = tf.keras.Input(shape=(len_seq_max,))
+    embedding_layer = TokenAndPositionEmbedding(len_seq_max, 25, 32)
+    x = embedding_layer(Transformer_input)
+    transformer_block = TransformerBlock(32, 8, 32)
+    x = transformer_block(x)
+    x = layers.GlobalAveragePooling1D()(x)
+    x = layers.Dropout(0.1)(x)
+    x = layers.Dense(20, activation="relu")(x)
+    x = layers.Dropout(0.1)(x)
+    Transformer_output = layers.Dense(256, activation="relu")(x)
+
+    lstm_input = tf.keras.Input(shape=(1, 511), name="lstm_input")
+    x = layers.LSTM(128, return_sequences=False)(lstm_input)
+    lstm_output = layers.Dense(1, activation="relu")(x)
+    output = layers.concatenate([Transformer_output, lstm_output])
+    outputss = layers.Dense(1, activation="sigmoid")(output)
+    model = Model(
+        inputs={'Transformer_input': Transformer_input, 'lstm_input': lstm_input},
+        outputs=outputss,
+    )
+
+    model.load_weights(filepath="AMP_818.h5", by_name=False, skip_mismatch=False, options=None)
+    proba = model.predict([test2,test1])
+    proba0 = (1-proba)*100
+    proba1 = (proba)*100
+    proba = transfer_label_from_prob(proba)
+
+    f = open (r'output.txt','a')
+    for i in range(len(proba)):
+        if proba[i]==0:
+            print(sequence[i], "Non-AMP", "%.3f%%"%proba0[i],file = f)
+        else:
+            print(sequence[i], "AMP", "%.3f%%"%proba1[i],file = f)
+    f.close()
+    return 'output.txt'
+     
+
+iface = gr.Interface(fn=ACP_DL,
+                     inputs = [gr.File(label="input fasta")], 
+                     outputs= gr.File(label="download txt"))
+iface.launch()