MLP_model.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{'data': tensor([[ 0.0224,  0.0143, -0.0020,  ..., -0.0495, -0.1222, -0.0426],\n",
      "        [ 0.0247,  0.0146,  0.0007,  ..., -0.0586, -0.1254, -0.0454],\n",
      "        [ 0.0040, -0.0263, -0.0179,  ..., -0.0267, -0.1128, -0.0810],\n",
      "        ...,\n",
      "        [ 0.0121, -0.0161,  0.0114,  ..., -0.0207, -0.1261, -0.0674],\n",
      "        [ 0.0416,  0.0167,  0.0156,  ..., -0.0460, -0.1191, -0.0480],\n",
      "        [ 0.0365, -0.0047,  0.0034,  ..., -0.0487, -0.1371, -0.0620]]), 'proteins': ['407', '4011', 'L6', '3061', 'L141', 'CD7', 'CE8', '3B11', '4B1', 'B9', 'L1', 'L19', 'L24', 'P13', 'L39', 'L77', 'L149', 'L168', '4081', 'L28', 'L150', 'CE10', 'P10', 'L61', 'L170', 'CB6', 'CE7', '203', 'P1', 'P3', 'L15', 'L116', 'L133', 'L174', 'TfR_1', 'TfR_2', 'P6', 'n501', 'L2', 'L7', 'L16', 'L88', 'L91', 'L93', 'L175', 'P4', 'P5', 'P11', 'P18', 'P19', 'P20', 'P21', 'C8', 'R5', 'L14', 'L23', 'L45', 'L70', 'L143', 'L151', 'L177', 'HSA_5', '301', 'P2', 'P15', 'L20', 'L25', 'L32', 'L37', 'L57', 'L72', 'L86', 'L136', 'L142', 'L147', 'L148', '3B2', '3B7', 'HSA_3', 'HSA_4', '406', 'P7', 'P8', 'P9', 'L73', 'L125', 'L134', 'L158', 'L160', 'L171', '3023', '3063', 'L38', 'L50', '201', 'P12', 'P16', 'L154', '202', 'P17', 'L85', 'L138', 'P14', '4023', 'L22']}\n",
      "torch.Size([105, 2560])\n"
     ]
    }
   ],
   "source": [
    "import torch\n",
    "tensor = torch.load('train_esm.pt')\n",
    "print(tensor)\n",
    "embed=tensor[\"data\"]\n",
    "print(embed.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "from pandas import read_csv\n",
    "data=read_csv('../Yinglab_Protein_Data/UDAB_production_20240521.csv')\n",
    "#取出production_classify列，如果是A就是1，如果是B就是0\n",
    "label = data['production_classify'].map({'A': 1, 'B': 0})"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "torch.Size([105, 2560])\n",
      "torch.FloatTensor\n",
      "torch.Size([105])\n",
      "torch.LongTensor\n"
     ]
    }
   ],
   "source": [
    "all_data = torch.load('train_esm.pt')[\"data\"]\n",
    "all_label=label\n",
    "#把label转换成tensor\n",
    "all_label = torch.tensor(all_label.values, dtype=torch.long)\n",
    "print(all_data.shape)\n",
    "print(all_data.type())\n",
    "print(all_label.shape)\n",
    "print(all_label.type())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [],
   "source": [
    "#将数据集分为训练集和测试集\n",
    "from sklearn.model_selection import train_test_split\n",
    "train_data, test_data, train_label, test_label = train_test_split(all_data, all_label, test_size=0.05, random_state=0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch [1/15], Loss: 0.6942394375801086\n",
      "Epoch [2/15], Loss: 0.6905669569969177\n",
      "Epoch [3/15], Loss: 0.6878826022148132\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch [4/15], Loss: 0.6856376528739929\n",
      "Epoch [5/15], Loss: 0.6837663054466248\n",
      "Epoch [6/15], Loss: 0.6822030544281006\n",
      "Epoch [7/15], Loss: 0.6808946132659912\n",
      "Epoch [8/15], Loss: 0.6797977685928345\n",
      "Epoch [9/15], Loss: 0.6789182424545288\n",
      "Epoch [10/15], Loss: 0.6782863736152649\n",
      "Epoch [11/15], Loss: 0.6779224872589111\n",
      "Epoch [12/15], Loss: 0.6778190732002258\n",
      "Epoch [13/15], Loss: 0.677920937538147\n",
      "Epoch [14/15], Loss: 0.6781055927276611\n",
      "Epoch [15/15], Loss: 0.6782318949699402\n"
     ]
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "import torch\n",
    "import torch.nn as nn\n",
    "import torch.optim as optim\n",
    "\n",
    "class MLP(nn.Module):\n",
    "    def __init__(self, input_size, hidden_size_1,hidden_size_2,hidden_size_3,output_size):\n",
    "        super(MLP, self).__init__()\n",
    "        self.hidden_1 = nn.Linear(input_size, hidden_size_1)\n",
    "        self.relu = nn.ReLU()\n",
    "        self.hidden_2 = nn.Linear(hidden_size_1, hidden_size_2)\n",
    "        self.hidden_3 = nn.Linear(hidden_size_2, hidden_size_3)\n",
    "        self.output = nn.Linear(hidden_size_3, output_size)\n",
    "        \n",
    "    def forward(self, x):\n",
    "        x = self.hidden_1(x)\n",
    "        x = self.relu(x)\n",
    "        x = self.hidden_2(x)\n",
    "        x = self.relu(x)\n",
    "        x = self.hidden_3(x)\n",
    "        x = self.relu(x)\n",
    "        x = self.output(x)\n",
    "        return x\n",
    "\n",
    "input_size = 2560\n",
    "hidden_size_1 = 1120\n",
    "hidden_size_2 = 560\n",
    "hidden_size_3 = 280\n",
    "output_size = 2\n",
    "model = MLP(input_size, hidden_size_1,hidden_size_2, hidden_size_3,output_size)\n",
    "\n",
    "criterion = nn.CrossEntropyLoss()\n",
    "optimizer = optim.Adam(model.parameters(), lr=0.0001)\n",
    "\n",
    "input_data = train_data\n",
    "target_data = train_label\n",
    "\n",
    "num_epochs = 15\n",
    "losslist = []\n",
    "for epoch in range(num_epochs):\n",
    "    output = model(input_data)\n",
    "    loss = criterion(output, target_data)\n",
    "    losslist.append(loss.item())\n",
    "    # 反向传播和优化\n",
    "    optimizer.zero_grad()\n",
    "    loss.backward()\n",
    "    optimizer.step()\n",
    "    \n",
    "    print(f'Epoch [{epoch+1}/{num_epochs}], Loss: {loss.item()}')\n",
    "#画图，横坐标是epoch，纵坐标是loss\n",
    "import matplotlib.pyplot as plt\n",
    "plt.plot(losslist)\n",
    "\n",
    "\n",
    "# 测试模型\n",
    "test_input = test_data\n",
    "with torch.no_grad():\n",
    "    output = model(test_input)\n",
    "    predicted_classes = torch.argmax(output, dim=1)\n",
    "    # 计算准确率\n",
    "    accuracy = (predicted_classes == test_label).sum() / test_label.size(0) \n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "6\n",
      "the correction rate is: 66.67%\n"
     ]
    }
   ],
   "source": [
    "# 测试模型\n",
    "test_input = test_data\n",
    "with torch.no_grad():\n",
    "    output = model(test_input)\n",
    "    predicted_classes = torch.argmax(output, dim=1)\n",
    "    # 计算准确率\n",
    "accuracy = (predicted_classes == test_label).sum() / test_label.size(0)\n",
    "print(test_label.size(0))\n",
    "print(f\"the correction rate is: {accuracy:.2%}\")"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "pytorch2",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.9.18"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}