app.py

import gradio as gr
import pandas as pd

import io

from Bio import SeqIO
from tape import ProteinBertModel, ProteinBertConfig, TAPETokenizer  # type: ignore
from tape.models import modeling_bert
import numpy as np
import torch


tokenizer = TAPETokenizer(vocab='iupac')
config=modeling_bert.ProteinBertConfig(num_hidden_layers=5,num_attention_heads=8,hidden_size=400)

bert_model = torch.load('models/transformer1500_95p_500.pt')
class_model=torch.load('models/down_model_500_kfold1.pt')

bert_model=bert_model.module
bert_model=bert_model.to("cpu")
bert_model=bert_model.eval()


def func(name):

    translation_table = str.maketrans("", "", " \t\n\r\f\v")
    name = name.translate(translation_table)
    token_ids = torch.tensor([tokenizer.encode(name)])
    token_ids = token_ids
    bert_output = bert_model(token_ids)
    class_output=class_model(bert_output[1])
    class_output = torch.softmax(class_output, dim=1)
    cluster = torch.argmax(class_output, dim=1) + 1
    cluster=cluster.item()

    return "cluster "+str(cluster)


def func_mult(name):
    sequence_list = name.split("\n")
    sequence_list = [s.strip() for s in sequence_list]
    sequence_list = [x for x in sequence_list if x]  # 列表推导式
    output=[]
    for i in range(0, len(sequence_list), 1):
        output.append(func(sequence_list[i]))
    result = "\n".join(output)
    return result


def read_fasta_file(file_path):
    sequences = []
    for seq_record in SeqIO.parse(file_path, "fasta"):
        sequences.append(str(seq_record.seq))
    return sequences


def func_file(file_path):
    sequence_list = read_fasta_file(file_path)
    output=[]
    for i in range(0, len(sequence_list), 1):
        output.append(func(sequence_list[i]))
    result = "\n".join(output)
    return result


def upload_file(files):
    file_paths = [file.name for file in files]
    return file_paths[0]


def save_to_txt(data):
    # 写入数据到 TXT 文件
    file_name="output.txt"
    with open(file_name, mode='w') as file:
        file.write(data)

    # 返回文件路径
    return file_name


css = """
.gradio-container {background-color: #EDEFF7}
.button {background-color: #515D90; color:#FFFFFF}
.feedback {font-size: 36px}
"""

with gr.Blocks(css=css, title="GH29 Prediction", theme=gr.themes.Soft()) as demo:
    gr.Markdown("GH29 Prediction", elem_classes="feedback")

    # 创建一个包含Markdown说明的示例块

    with gr.Tab("Single sequence input"):
        with gr.Row():
            single_input = gr.Textbox(lines=10, placeholder="Please input sequence data (note: do not input fasta data)", label="Input")
            single_output = gr.Textbox(lines=10, label="Output", show_copy_button=True)
        single_button = gr.Button("Predict", elem_classes="button")

    with gr.Tab("Multiple sequence input"):
        multiple_input = gr.Textbox(lines=10, placeholder="Please enter multiple sequence data separated by line breaks (do not enter fasta data)", label="Input")
        multiple_button = gr.Button("Predict", elem_classes="button")
        multiple_output = gr.Textbox(lines=10, label="Output", show_copy_button=True)

    with gr.Tab("FASTA input"):
        with gr.Row():
            file_upload = gr.File(label="Fasta File", interactive=False, scale=2)
            file_output_textbox = gr.Textbox(lines=15, label="Output", scale=3, container=True, autoscroll=True, show_copy_button=True)
            file_output_file = gr.File(label="Output File", scale=2)
        with gr.Row():
            upload_button = gr.UploadButton("Click to Upload a File", file_types=["fasta"], scale=2, size="sm", file_count="multiple")
            upload_button.upload(upload_file, upload_button, file_upload)
            file_button = gr.Button("Predict", scale=3, size="lg", elem_classes="button")
            file_button_GenerateFile = gr.Button("Save to file", scale=2, size="sm")

    single_button.click(func, inputs=single_input, outputs=single_output)
    multiple_button.click(func_mult, inputs=multiple_input, outputs=multiple_output)
    file_button.click(func_file, inputs=file_upload, outputs=file_output_textbox)
    file_button_GenerateFile.click(save_to_txt, inputs=file_output_textbox, outputs=file_output_file)


demo.launch(share=True)