2023-12-18 15:45:05 +01:00
{
"cells": [
{
"cell_type": "markdown",
"id": "8bc02404-8cd1-46d9-8237-2d035ebb3e79",
"metadata": {},
"source": [
"# **[Project] Cancer Subtype Classification**"
]
},
{
"cell_type": "markdown",
"id": "0c5076f4",
"metadata": {},
"source": [
"# Introduction"
]
},
{
"cell_type": "markdown",
"id": "8a599748",
"metadata": {},
"source": [
"The [TCGA Kidney Cancers Dataset](https://archive.ics.uci.edu/dataset/892/tcga+kidney+cancers) is a bulk RNA-seq dataset that contains transcriptome profiles (i.e., gene expression quantification data) of patients diagnosed with three different subtypes of kidney cancers.\n",
"This dataset can be used to make predictions about the specific subtype of kidney cancers given the normalized transcriptome profile data.\n",
"\n",
"The normalized transcriptome profile data is given as **TPM** and **FPKM** for each gene.\n",
"\n",
"> TPM (Transcripts Per Million) and FPKM (Fragments Per Kilobase Million) are two common methods for quantifying gene expression in RNA sequencing data.\n",
"> They both aim to account for the differences in sequencing depth and transcript length when estimating gene expression levels.\n",
">\n",
"> **TPM** (Transcripts Per Million):\n",
"> - TPM is a measure of gene expression that normalizes for both library size (sequencing depth) and transcript length.\n",
"> - The main idea behind TPM is to express the abundance of a transcript relative to the total number of transcripts in a sample, scaled to one million.\n",
">\n",
"> **FPKM** (Fragments Per Kilobase Million):\n",
"> - FPKM is another method for quantifying gene expression, which is commonly used in older RNA-seq analysis pipelines. It's similar in concept to TPM but differs in the way it's calculated.\n",
"> - FPKM also normalizes for library size and transcript length, but it measures gene expression as the number of fragments (i.e., reads) per kilobase of exon model per million reads.\n",
">\n",
"> TPM is generally considered more robust to variations in library size, making it a preferred choice in many modern RNA-seq analysis workflows.\n",
"\n",
"We provide one dataset for each kidney cancer subtype:\n",
"\n",
"- [TCGA-KICH](https://portal.gdc.cancer.gov/projects/TCGA-KICH): kidney chromophobe (renal clear cell carcinoma)\n",
"- [TCGA-KIRC](https://portal.gdc.cancer.gov/projects/TCGA-KIRC): kidney renal clear cell carcinoma\n",
"- [TCGA-KIRP](https://portal.gdc.cancer.gov/projects/TCGA-KIRP): kidney renal papillary cell carcinoma\n",
"\n",
"> This and _much_ more data is openly available on the [NCI Genomic Data Commons (GDC) Data Portal](https://portal.gdc.cancer.gov/)."
]
},
{
"cell_type": "markdown",
"id": "16712787",
"metadata": {},
"source": [
"# Data access"
]
},
{
"cell_type": "markdown",
"id": "6421ef6c",
"metadata": {},
"source": [
"There are two ways to access the data: via the TNT homepage or the GDC Data Portal."
]
},
{
"cell_type": "markdown",
"id": "b977e8b8",
"metadata": {},
"source": [
"## Download from the TNT homepage (_recommended_)"
]
},
{
"cell_type": "markdown",
"id": "800fa7bd",
"metadata": {},
"source": [
"The download from the TNT homepage is straightforward:"
]
},
{
"cell_type": "code",
"execution_count": 1,
"id": "dda97b16",
"metadata": {
"tags": []
},
"outputs": [
{
2023-12-29 11:06:25 +01:00
"ename": "SyntaxError",
"evalue": "invalid syntax (2666948873.py, line 6)",
"output_type": "error",
"traceback": [
"\u001b[0;36m Cell \u001b[0;32mIn[1], line 6\u001b[0;36m\u001b[0m\n\u001b[0;31m from IPython.display import clear_output(wait=True)\u001b[0m\n\u001b[0m ^\u001b[0m\n\u001b[0;31mSyntaxError\u001b[0m\u001b[0;31m:\u001b[0m invalid syntax\n"
2023-12-18 15:45:05 +01:00
]
}
],
"source": [
"! wget http://www.tnt.uni-hannover.de/edu/vorlesungen/AMLG/data/project-cancer-classification.tar.gz\n",
"! tar -xzvf project-cancer-classification.tar.gz\n",
"! mv -v project-cancer-classification/ data/\n",
"! rm -v project-cancer-classification.tar.gz"
]
},
{
"cell_type": "markdown",
"id": "bc2db880",
"metadata": {},
"source": [
"In the `data/` folder you will now find many files in the [TSV format](https://en.wikipedia.org/wiki/Tab-separated_values) ([CSV](https://en.wikipedia.org/wiki/Comma-separated_values)-like with tabs as delimiter) containing the normalized transcriptome profile data.\n",
"\n",
"To start, you can read a TSV file into a [pandas](https://pandas.pydata.org) [`DataFrame`](pandas dataframe to dict) using the [`pandas.read_csv()`](https://pandas.pydata.org/docs/reference/api/pandas.read_csv.html#pandas-read-csv) function with the `sep` parameter set to `\\t`:"
]
},
2023-12-29 11:06:25 +01:00
{
"cell_type": "markdown",
"id": "ed50d396-fe33-47a7-ad19-8eb975ef0fa5",
"metadata": {},
"source": [
"## Lesen der DNA-Sequenz Dateien und speichern in einer Datei"
]
},
2023-12-18 15:45:05 +01:00
{
"cell_type": "code",
2024-01-04 11:32:45 +01:00
"execution_count": 1,
2023-12-18 15:45:05 +01:00
"id": "2adae4ff",
"metadata": {
"tags": []
},
2024-01-04 11:32:45 +01:00
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"Es wurden 1034 Dateien eingelesen.\n"
]
}
],
2023-12-18 15:45:05 +01:00
"source": [
2023-12-29 11:06:25 +01:00
"import numpy as np\n",
2023-12-18 15:45:05 +01:00
"import pandas as pd\n",
"import pickle\n",
"\n",
"\n",
"import os\n",
"#'./data/tcga-kirp-geq'\n",
"\n",
"labels = [\"kirp\", \"kirc\", \"kich\"] # Setzen Sie hier Ihren Ordnerpfad ein\n",
2023-12-29 11:06:25 +01:00
"n_files = 0\n",
"y = list()\n",
"x = list()\n",
2023-12-18 15:45:05 +01:00
"\n",
"rick = list()\n",
2023-12-29 11:06:25 +01:00
"data = []\n",
"\n",
2023-12-18 15:45:05 +01:00
"for l in labels:\n",
" root_folder = f\"./data/tcga-{l}-geq\"\n",
" for root, dirs, files in os.walk(root_folder):\n",
" for file in files:\n",
" if file.endswith('.tsv'):\n",
2023-12-29 11:06:25 +01:00
" n_files += 1\n",
2023-12-18 15:45:05 +01:00
" # Vollständiger Pfad zur Datei\n",
" file_path = os.path.join(root, file)\n",
" # Hier können Sie etwas mit der Datei machen, z.B. einlesen\n",
" df = pd.read_csv(filepath_or_buffer=file_path, sep=\"\\t\", header=1)\n",
" df = df['tpm_unstranded']\n",
"\n",
" df = df[4:]\n",
" df = np.array(df)\n",
" rick.append(df)\n",
2023-12-29 11:06:25 +01:00
" \n",
" data.append([df, l])\n",
2023-12-18 15:45:05 +01:00
"\n",
2023-12-29 11:06:25 +01:00
"print(f\"Es wurden {n_files} Dateien eingelesen.\")\n",
2023-12-18 15:45:05 +01:00
"#tsv_file_path = \"data/tcga-kich-geq/0ba21ef5-0829-422e-a674-d3817498c333/4868e8fc-e045-475a-a81d-ef43eabb7066.rna_seq.augmented_star_gene_counts.tsv\"\n",
"\n",
"# Read the TSV file into a DataFrame\n",
"#df = pd.read_csv(filepath_or_buffer=tsv_file_path, sep=\"\\t\", header=1)\n",
"\n",
"# Display the first few rows of the DataFrame\n",
"#print(df.head(n=20))\n",
2024-01-04 11:04:09 +01:00
"#rick = np.array(rick)\n",
2023-12-18 15:45:05 +01:00
"\n",
"# Speichern der 'kirp' Liste in einer Pickle-Datei\n",
2024-01-04 11:04:09 +01:00
"#with open('rick.pickle', 'wb') as f:\n",
"# pickle.dump(rick, f)\n"
2023-12-18 15:45:05 +01:00
]
},
{
"cell_type": "code",
2024-01-04 11:32:45 +01:00
"execution_count": 2,
2023-12-29 11:06:25 +01:00
"id": "dfe4f964-6068-46da-8103-194525086f01",
"metadata": {
"tags": []
},
2024-01-04 11:32:45 +01:00
"outputs": [
{
"data": {
"text/html": [
"<div>\n",
"<style scoped>\n",
" .dataframe tbody tr th:only-of-type {\n",
" vertical-align: middle;\n",
" }\n",
"\n",
" .dataframe tbody tr th {\n",
" vertical-align: top;\n",
" }\n",
"\n",
" .dataframe thead th {\n",
" text-align: right;\n",
" }\n",
"</style>\n",
"<table border=\"1\" class=\"dataframe\">\n",
" <thead>\n",
" <tr style=\"text-align: right;\">\n",
" <th></th>\n",
" <th>genome_frequencies</th>\n",
" <th>cancer_type</th>\n",
" </tr>\n",
" </thead>\n",
" <tbody>\n",
" <tr>\n",
" <th>0</th>\n",
" <td>[20.331, 0.0, 25.1806, 1.1301, 0.4836, 7.3269,...</td>\n",
" <td>kirp</td>\n",
" </tr>\n",
" <tr>\n",
" <th>1</th>\n",
" <td>[37.0405, 0.5002, 77.4246, 4.2188, 1.0408, 29....</td>\n",
" <td>kirp</td>\n",
" </tr>\n",
" <tr>\n",
" <th>2</th>\n",
" <td>[45.4456, 0.0903, 74.9545, 4.843, 1.5188, 11.8...</td>\n",
" <td>kirp</td>\n",
" </tr>\n",
" <tr>\n",
" <th>3</th>\n",
" <td>[15.2345, 0.3393, 62.0003, 2.4412, 0.932, 2.66...</td>\n",
" <td>kirp</td>\n",
" </tr>\n",
" <tr>\n",
" <th>4</th>\n",
" <td>[35.0709, 0.2333, 62.8022, 2.8872, 1.0547, 18....</td>\n",
" <td>kirp</td>\n",
" </tr>\n",
" </tbody>\n",
"</table>\n",
"</div>"
],
"text/plain": [
" genome_frequencies cancer_type\n",
"0 [20.331, 0.0, 25.1806, 1.1301, 0.4836, 7.3269,... kirp\n",
"1 [37.0405, 0.5002, 77.4246, 4.2188, 1.0408, 29.... kirp\n",
"2 [45.4456, 0.0903, 74.9545, 4.843, 1.5188, 11.8... kirp\n",
"3 [15.2345, 0.3393, 62.0003, 2.4412, 0.932, 2.66... kirp\n",
"4 [35.0709, 0.2333, 62.8022, 2.8872, 1.0547, 18.... kirp"
]
},
"execution_count": 2,
"metadata": {},
"output_type": "execute_result"
}
],
2023-12-29 11:06:25 +01:00
"source": [
"data_Frame = pd.DataFrame(data, columns=[\"genome_frequencies\", \"cancer_type\"])\n",
"data_Frame.head()"
]
},
2024-01-04 11:04:09 +01:00
{
"cell_type": "code",
2024-01-04 11:32:45 +01:00
"execution_count": 3,
2024-01-04 11:04:09 +01:00
"id": "0f5cc92a-4485-4184-845e-116ea9a9776d",
2024-01-04 11:32:45 +01:00
"metadata": {
"tags": []
},
2024-01-04 11:04:09 +01:00
"outputs": [],
"source": [
2024-01-04 11:32:45 +01:00
"# Speichern der Daten in einer lokalen Datei\n",
2024-01-04 11:04:09 +01:00
"with open('rick.pickle', 'wb') as f:\n",
2024-01-04 11:32:45 +01:00
" pickle.dump(data_Frame, f)"
]
},
{
"cell_type": "code",
"execution_count": 2,
"id": "b7b79958-baba-4630-9def-cf47afe43d9f",
"metadata": {
"tags": []
},
"outputs": [],
"source": [
"import pickle\n",
"\n",
"# Laden der 'kirp' Liste aus der Pickle-Datei\n",
"with open('rick.pickle', 'rb') as f:\n",
" data_Frame = pickle.load(f)"
]
},
{
"cell_type": "code",
"execution_count": 3,
"id": "f6608b92-8ace-4a52-a3dc-70c578e56f0d",
"metadata": {
"tags": []
},
"outputs": [
{
"data": {
"text/html": [
"<div>\n",
"<style scoped>\n",
" .dataframe tbody tr th:only-of-type {\n",
" vertical-align: middle;\n",
" }\n",
"\n",
" .dataframe tbody tr th {\n",
" vertical-align: top;\n",
" }\n",
"\n",
" .dataframe thead th {\n",
" text-align: right;\n",
" }\n",
"</style>\n",
"<table border=\"1\" class=\"dataframe\">\n",
" <thead>\n",
" <tr style=\"text-align: right;\">\n",
" <th></th>\n",
" <th>genome_frequencies</th>\n",
" <th>cancer_type</th>\n",
" </tr>\n",
" </thead>\n",
" <tbody>\n",
" <tr>\n",
" <th>0</th>\n",
" <td>[20.331, 0.0, 25.1806, 1.1301, 0.4836, 7.3269,...</td>\n",
" <td>kirp</td>\n",
" </tr>\n",
" <tr>\n",
" <th>1</th>\n",
" <td>[37.0405, 0.5002, 77.4246, 4.2188, 1.0408, 29....</td>\n",
" <td>kirp</td>\n",
" </tr>\n",
" <tr>\n",
" <th>2</th>\n",
" <td>[45.4456, 0.0903, 74.9545, 4.843, 1.5188, 11.8...</td>\n",
" <td>kirp</td>\n",
" </tr>\n",
" <tr>\n",
" <th>3</th>\n",
" <td>[15.2345, 0.3393, 62.0003, 2.4412, 0.932, 2.66...</td>\n",
" <td>kirp</td>\n",
" </tr>\n",
" <tr>\n",
" <th>4</th>\n",
" <td>[35.0709, 0.2333, 62.8022, 2.8872, 1.0547, 18....</td>\n",
" <td>kirp</td>\n",
" </tr>\n",
" </tbody>\n",
"</table>\n",
"</div>"
],
"text/plain": [
" genome_frequencies cancer_type\n",
"0 [20.331, 0.0, 25.1806, 1.1301, 0.4836, 7.3269,... kirp\n",
"1 [37.0405, 0.5002, 77.4246, 4.2188, 1.0408, 29.... kirp\n",
"2 [45.4456, 0.0903, 74.9545, 4.843, 1.5188, 11.8... kirp\n",
"3 [15.2345, 0.3393, 62.0003, 2.4412, 0.932, 2.66... kirp\n",
"4 [35.0709, 0.2333, 62.8022, 2.8872, 1.0547, 18.... kirp"
]
},
"execution_count": 3,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"data_Frame.head()"
2024-01-04 11:04:09 +01:00
]
},
2023-12-29 11:06:25 +01:00
{
"cell_type": "markdown",
"id": "c60cbf60-d904-4ee0-8f70-588bb109368b",
"metadata": {},
"source": [
"# Data preprocessing"
]
},
{
"cell_type": "markdown",
"id": "583e39c8-13ba-422e-9c39-9cf1c8d63d5b",
"metadata": {},
"source": [
"## Training set & validation set"
]
},
{
"cell_type": "code",
2024-01-04 11:32:45 +01:00
"execution_count": 4,
2023-12-29 11:06:25 +01:00
"id": "38695a70-86e9-4dd0-b622-33e3762372eb",
"metadata": {
"tags": []
},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"DataSet shape: (1034, 2)\n",
"Training set\n",
"------------\n",
"Dataframe shape: (827, 2)\n",
"Dataframe head:\n",
" genome_frequencies cancer_type\n",
"518 [25.0645, 0.1125, 56.3997, 3.3108, 1.6061, 12.... kirc\n",
"355 [32.6449, 2.1789, 63.4954, 6.3228, 2.109, 40.9... kirc\n",
"528 [46.024, 0.0, 85.8077, 7.2567, 2.1301, 9.6509,... kirc\n",
"445 [153.0064, 1.6403, 99.3267, 7.3736, 1.3668, 10... kirc\n",
"986 [65.5167, 18.2363, 77.2126, 5.0375, 2.4628, 21... kich\n",
"\n",
"Validation set\n",
"--------------\n",
"Dataframe shape: (207, 2)\n",
"Dataframe head:\n",
" genome_frequencies cancer_type\n",
"294 [50.8994, 0.4635, 131.5049, 5.7193, 3.103, 15.... kirp\n",
"453 [35.857, 0.1018, 94.5681, 5.2997, 1.9388, 17.6... kirc\n",
"638 [11.3865, 0.2313, 28.5961, 3.0169, 0.7851, 8.2... kirc\n",
"139 [41.6119, 0.2207, 55.4377, 4.4395, 0.884, 3.56... kirp\n",
"539 [63.1646, 18.8107, 63.2703, 4.6696, 0.9466, 5.... kirc\n"
]
}
],
"source": [
"import os\n",
"from sklearn.model_selection import train_test_split\n",
"\n",
"train_df, val_df = train_test_split(data_Frame, train_size=0.8, random_state=42)\n",
"\n",
"print(f\"DataSet shape: {data_Frame.shape}\")\n",
"print(f\"Training set{os.linesep}------------\")\n",
"print(f\"Dataframe shape: {train_df.shape}\")\n",
"print(f\"Dataframe head:{os.linesep}{train_df.head()}\")\n",
"print(\"\")\n",
"print(f\"Validation set{os.linesep}--------------\")\n",
"print(f\"Dataframe shape: {val_df.shape}\")\n",
"print(f\"Dataframe head:{os.linesep}{val_df.head()}\")"
]
},
{
"cell_type": "markdown",
"id": "4903244b-548f-4672-967d-1c62825b6fce",
"metadata": {},
"source": [
"## Building a custom PyTorch dataset"
]
},
{
"cell_type": "markdown",
"id": "7e333251-c4e7-41f0-a086-12a3d95b723f",
"metadata": {},
"source": [
"## Öffnen der Datei mit den Gesammelten Sequenzen"
]
},
{
"cell_type": "code",
2024-01-04 11:32:45 +01:00
"execution_count": 5,
2023-12-29 11:06:25 +01:00
"id": "e2f78725-cda6-4e8d-9029-a4a31f6f9ab7",
"metadata": {
"tags": []
},
"outputs": [],
"source": [
"from torch.utils.data import Dataset\n",
"import torch\n",
"import pandas as pd\n",
"from sklearn.preprocessing import LabelEncoder\n",
"\n",
"class GenomeDataset(Dataset):\n",
" def __init__(self, dataframe):\n",
" self.dataframe = dataframe\n",
"\n",
" # Umwandlung der Genome Frequenzen in Tensoren\n",
" self.genome_frequencies = torch.tensor(dataframe['genome_frequencies'].tolist(), dtype=torch.float32)\n",
"\n",
" # Umwandlung der Krebsarten in numerische Werte\n",
" self.label_encoder = LabelEncoder()\n",
" self.cancer_types = torch.tensor(self.label_encoder.fit_transform(dataframe['cancer_type']), dtype=torch.long)\n",
"\n",
" def __getitem__(self, index):\n",
" # Rückgabe eines Tupels aus Genome Frequenzen und dem entsprechenden Krebstyp\n",
" return self.genome_frequencies[index], self.cancer_types[index]\n",
"\n",
" def __len__(self):\n",
" return len(self.dataframe)\n"
]
},
{
"cell_type": "code",
2024-01-04 11:32:45 +01:00
"execution_count": 6,
2023-12-29 11:06:25 +01:00
"id": "aaa2c50c-c79e-4bca-812f-1a06c9f485d5",
"metadata": {
"tags": []
},
2024-01-04 11:32:45 +01:00
"outputs": [
{
"name": "stderr",
"output_type": "stream",
"text": [
"/tmp/ipykernel_343/2483914749.py:11: UserWarning: Creating a tensor from a list of numpy.ndarrays is extremely slow. Please consider converting the list to a single numpy.ndarray with numpy.array() before converting to a tensor. (Triggered internally at ../torch/csrc/utils/tensor_new.cpp:245.)\n",
" self.genome_frequencies = torch.tensor(dataframe['genome_frequencies'].tolist(), dtype=torch.float32)\n"
]
}
],
2023-12-29 11:06:25 +01:00
"source": [
"# Beispielhafte Verwendung\n",
"# Angenommen, df_train und df_valid sind Ihre Trainings- und Validierungsdaten\n",
"train_dataset = GenomeDataset(train_df)\n",
"valid_dataset = GenomeDataset(val_df)"
]
},
{
"cell_type": "code",
2024-01-04 11:32:45 +01:00
"execution_count": 7,
2023-12-29 11:06:25 +01:00
"id": "a7fb59af-bd06-42d4-acce-03266a85bf36",
"metadata": {
"tags": []
},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"Genome frequency from dataframe:\n",
"[2.50645e+01 1.12500e-01 5.63997e+01 ... 0.00000e+00 1.29000e-02\n",
" 2.47100e-01]\n",
"\n",
"Cancer type from dataframe: kirc\n",
"\n",
"Genome frequency from dataset:\n",
"tensor([2.5065e+01, 1.1250e-01, 5.6400e+01, ..., 0.0000e+00, 1.2900e-02,\n",
" 2.4710e-01])\n",
"\n",
"Cancer type from dataset: 1\n"
]
}
],
"source": [
"# Inspect the first item from the training dataframe\n",
"train_df_head = train_df.head(n=1)\n",
"train_df_genome_frequence =train_df_head.iloc[0][\"genome_frequencies\"]\n",
"train_df_cancer_type = train_df_head.iloc[0][\"cancer_type\"]\n",
"print(f\"Genome frequency from dataframe:{os.linesep}{train_df_genome_frequence}{os.linesep}\")\n",
"print(f\"Cancer type from dataframe: {train_df_cancer_type}{os.linesep}\")\n",
"\n",
"# Inspect the first item from the training dataset\n",
"datapoint_features, datapoint_label = train_dataset[0]\n",
"print(f\"Genome frequency from dataset:{os.linesep}{datapoint_features}{os.linesep}\")\n",
"print(f\"Cancer type from dataset: {datapoint_label}\")"
]
},
{
"cell_type": "markdown",
"id": "9199fdeb-0d48-44c2-8bec-db2a7d7cbd4d",
"metadata": {},
"source": [
"# Neuronales Netz Definition"
]
},
2024-01-04 11:32:45 +01:00
{
"cell_type": "markdown",
"id": "e53132b9-6222-4739-be49-7628e5a37709",
"metadata": {},
"source": [
"### Simples Neuronales Netz"
]
},
2023-12-29 11:06:25 +01:00
{
"cell_type": "code",
2024-01-04 11:32:45 +01:00
"execution_count": 8,
2023-12-29 11:06:25 +01:00
"id": "76b8eec8-d24b-4696-82bf-ebb286e7d1e7",
"metadata": {
"tags": []
},
"outputs": [],
"source": [
"import torch\n",
"import torch.nn as nn\n",
"import torch.optim as optim\n",
"from torch.utils.data import DataLoader\n",
"\n",
"# Definition des Modells\n",
"class SimpleNN(nn.Module):\n",
" def __init__(self, input_size, hidden_size, num_classes):\n",
" super(SimpleNN, self).__init__()\n",
" self.fc1 = nn.Linear(input_size, hidden_size)\n",
" self.relu = nn.ReLU()\n",
" self.fc2 = nn.Linear(hidden_size, num_classes)\n",
"\n",
" def forward(self, x):\n",
" out = self.fc1(x)\n",
" out = self.relu(out)\n",
" out = self.fc2(out)\n",
" return out"
]
},
2024-01-04 11:32:45 +01:00
{
"cell_type": "markdown",
"id": "e2e9e0dd-3d4f-4999-9e65-704266d5e4a2",
"metadata": {
"tags": []
},
"source": [
"### Komplexes Neuronales Netz"
]
},
{
"cell_type": "code",
"execution_count": 32,
"id": "944d463e-12ed-4447-8587-ee9c60ce3eb6",
"metadata": {
"tags": []
},
"outputs": [],
"source": [
"import torch\n",
"import torch.nn as nn\n",
"import torch.optim as optim\n",
"from torch.utils.data import DataLoader\n",
"\n",
"class ComplexNN(nn.Module):\n",
" def __init__(self, input_size, hidden_size, num_classes):\n",
" super(ComplexNN, self).__init__()\n",
" # Definieren der Schichten\n",
" self.fc1 = nn.Linear(input_size, 1024) # Eingabeschicht\n",
" self.fc2 = nn.Linear(1024, 512) # Versteckte Schicht\n",
" self.fc3 = nn.Linear(512, 256) # Weitere versteckte Schicht\n",
" self.fc4 = nn.Linear(256, num_classes) # Ausgabeschicht\n",
"\n",
" def forward(self, x):\n",
" # Definieren des Vorwärtsdurchlaufs\n",
" x = nn.ReLU(self.fc1(x))\n",
" x = nn.Dropout(p=0.5, inplace=False)\n",
" x = nn.ReLU(self.fc2(x))\n",
" x = nn.Dropout(p=0.5, inplace=False)\n",
" x = nn.ReLU(self.fc3(x))\n",
" x = torch.Sigmoid(self.fc4(x)) # Oder F.log_softmax für Mehrklassenklassifikation\n",
" return x"
]
},
2023-12-29 11:06:25 +01:00
{
"cell_type": "code",
2024-01-04 11:32:45 +01:00
"execution_count": 33,
2023-12-29 11:06:25 +01:00
"id": "60789428-7d6e-4737-a83a-1138f6a650f7",
"metadata": {
"tags": []
},
"outputs": [],
"source": [
"# Annahme: input_size ist die Länge Ihrer Genome-Frequenzen und num_classes ist die Anzahl der Krebsarten\n",
2024-01-04 11:32:45 +01:00
"#model = SimpleNN(input_size=60660, hidden_size=5000, num_classes=3)\n",
"model = ComplexNN(input_size=60660, hidden_size=5000, num_classes=3)\n",
2023-12-29 11:06:25 +01:00
"\n",
"# Daten-Loader\n",
"train_loader = DataLoader(dataset=train_dataset, batch_size=64, shuffle=True)\n",
"valid_loader = DataLoader(dataset=valid_dataset, batch_size=64, shuffle=False)"
]
},
{
"cell_type": "code",
2024-01-04 11:32:45 +01:00
"execution_count": 34,
2023-12-29 11:06:25 +01:00
"id": "de6e81de-0096-443a-a0b6-90cddecf5f88",
"metadata": {
"tags": []
},
"outputs": [],
"source": [
"# Verlustfunktion und Optimierer\n",
"criterion = nn.CrossEntropyLoss()\n",
"optimizer = optim.Adam(model.parameters(), lr=0.001)\n",
"num_epochs = 70"
]
},
{
"cell_type": "code",
2024-01-04 11:32:45 +01:00
"execution_count": 35,
2023-12-29 11:06:25 +01:00
"id": "a5deb2ed-c685-4d80-bc98-d6dd27334d82",
"metadata": {
"tags": []
},
"outputs": [
{
2024-01-04 11:32:45 +01:00
"ename": "TypeError",
"evalue": "linear(): argument 'input' (position 1) must be Tensor, not Dropout",
"output_type": "error",
"traceback": [
"\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
"\u001b[0;31mTypeError\u001b[0m Traceback (most recent call last)",
"Cell \u001b[0;32mIn[35], line 10\u001b[0m\n\u001b[1;32m 8\u001b[0m \u001b[38;5;28;01mfor\u001b[39;00m i, (inputs, labels) \u001b[38;5;129;01min\u001b[39;00m \u001b[38;5;28menumerate\u001b[39m(train_loader):\n\u001b[1;32m 9\u001b[0m optimizer\u001b[38;5;241m.\u001b[39mzero_grad()\n\u001b[0;32m---> 10\u001b[0m outputs \u001b[38;5;241m=\u001b[39m \u001b[43mmodel\u001b[49m\u001b[43m(\u001b[49m\u001b[43minputs\u001b[49m\u001b[43m)\u001b[49m\n\u001b[1;32m 11\u001b[0m loss \u001b[38;5;241m=\u001b[39m criterion(outputs, labels)\n\u001b[1;32m 12\u001b[0m loss\u001b[38;5;241m.\u001b[39mbackward()\n",
"File \u001b[0;32m/opt/conda/lib/python3.10/site-packages/torch/nn/modules/module.py:1501\u001b[0m, in \u001b[0;36mModule._call_impl\u001b[0;34m(self, *args, **kwargs)\u001b[0m\n\u001b[1;32m 1496\u001b[0m \u001b[38;5;66;03m# If we don't have any hooks, we want to skip the rest of the logic in\u001b[39;00m\n\u001b[1;32m 1497\u001b[0m \u001b[38;5;66;03m# this function, and just call forward.\u001b[39;00m\n\u001b[1;32m 1498\u001b[0m \u001b[38;5;28;01mif\u001b[39;00m \u001b[38;5;129;01mnot\u001b[39;00m (\u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39m_backward_hooks \u001b[38;5;129;01mor\u001b[39;00m \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39m_backward_pre_hooks \u001b[38;5;129;01mor\u001b[39;00m \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39m_forward_hooks \u001b[38;5;129;01mor\u001b[39;00m \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39m_forward_pre_hooks\n\u001b[1;32m 1499\u001b[0m \u001b[38;5;129;01mor\u001b[39;00m _global_backward_pre_hooks \u001b[38;5;129;01mor\u001b[39;00m _global_backward_hooks\n\u001b[1;32m 1500\u001b[0m \u001b[38;5;129;01mor\u001b[39;00m _global_forward_hooks \u001b[38;5;129;01mor\u001b[39;00m _global_forward_pre_hooks):\n\u001b[0;32m-> 1501\u001b[0m \u001b[38;5;28;01mreturn\u001b[39;00m \u001b[43mforward_call\u001b[49m\u001b[43m(\u001b[49m\u001b[38;5;241;43m*\u001b[39;49m\u001b[43margs\u001b[49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[38;5;241;43m*\u001b[39;49m\u001b[38;5;241;43m*\u001b[39;49m\u001b[43mkwargs\u001b[49m\u001b[43m)\u001b[49m\n\u001b[1;32m 1502\u001b[0m \u001b[38;5;66;03m# Do not call functions when jit is used\u001b[39;00m\n\u001b[1;32m 1503\u001b[0m full_backward_hooks, non_full_backward_hooks \u001b[38;5;241m=\u001b[39m [], []\n",
"Cell \u001b[0;32mIn[32], line 19\u001b[0m, in \u001b[0;36mComplexNN.forward\u001b[0;34m(self, x)\u001b[0m\n\u001b[1;32m 17\u001b[0m x \u001b[38;5;241m=\u001b[39m nn\u001b[38;5;241m.\u001b[39mReLU(\u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39mfc1(x))\n\u001b[1;32m 18\u001b[0m x \u001b[38;5;241m=\u001b[39m nn\u001b[38;5;241m.\u001b[39mDropout(p\u001b[38;5;241m=\u001b[39m\u001b[38;5;241m0.5\u001b[39m, inplace\u001b[38;5;241m=\u001b[39m\u001b[38;5;28;01mFalse\u001b[39;00m)\n\u001b[0;32m---> 19\u001b[0m x \u001b[38;5;241m=\u001b[39m nn\u001b[38;5;241m.\u001b[39mReLU(\u001b[38;5;28;43mself\u001b[39;49m\u001b[38;5;241;43m.\u001b[39;49m\u001b[43mfc2\u001b[49m\u001b[43m(\u001b[49m\u001b[43mx\u001b[49m\u001b[43m)\u001b[49m)\n\u001b[1;32m 20\u001b[0m x \u001b[38;5;241m=\u001b[39m nn\u001b[38;5;241m.\u001b[39mDropout(p\u001b[38;5;241m=\u001b[39m\u001b[38;5;241m0.5\u001b[39m, inplace\u001b[38;5;241m=\u001b[39m\u001b[38;5;28;01mFalse\u001b[39;00m)\n\u001b[1;32m 21\u001b[0m x \u001b[38;5;241m=\u001b[39m nn\u001b[38;5;241m.\u001b[39mReLU(\u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39mfc3(x))\n",
"File \u001b[0;32m/opt/conda/lib/python3.10/site-packages/torch/nn/modules/module.py:1501\u001b[0m, in \u001b[0;36mModule._call_impl\u001b[0;34m(self, *args, **kwargs)\u001b[0m\n\u001b[1;32m 1496\u001b[0m \u001b[38;5;66;03m# If we don't have any hooks, we want to skip the rest of the logic in\u001b[39;00m\n\u001b[1;32m 1497\u001b[0m \u001b[38;5;66;03m# this function, and just call forward.\u001b[39;00m\n\u001b[1;32m 1498\u001b[0m \u001b[38;5;28;01mif\u001b[39;00m \u001b[38;5;129;01mnot\u001b[39;00m (\u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39m_backward_hooks \u001b[38;5;129;01mor\u001b[39;00m \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39m_backward_pre_hooks \u001b[38;5;129;01mor\u001b[39;00m \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39m_forward_hooks \u001b[38;5;129;01mor\u001b[39;00m \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39m_forward_pre_hooks\n\u001b[1;32m 1499\u001b[0m \u001b[38;5;129;01mor\u001b[39;00m _global_backward_pre_hooks \u001b[38;5;129;01mor\u001b[39;00m _global_backward_hooks\n\u001b[1;32m 1500\u001b[0m \u001b[38;5;129;01mor\u001b[39;00m _global_forward_hooks \u001b[38;5;129;01mor\u001b[39;00m _global_forward_pre_hooks):\n\u001b[0;32m-> 1501\u001b[0m \u001b[38;5;28;01mreturn\u001b[39;00m \u001b[43mforward_call\u001b[49m\u001b[43m(\u001b[49m\u001b[38;5;241;43m*\u001b[39;49m\u001b[43margs\u001b[49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[38;5;241;43m*\u001b[39;49m\u001b[38;5;241;43m*\u001b[39;49m\u001b[43mkwargs\u001b[49m\u001b[43m)\u001b[49m\n\u001b[1;32m 1502\u001b[0m \u001b[38;5;66;03m# Do not call functions when jit is used\u001b[39;00m\n\u001b[1;32m 1503\u001b[0m full_backward_hooks, non_full_backward_hooks \u001b[38;5;241m=\u001b[39m [], []\n",
"File \u001b[0;32m/opt/conda/lib/python3.10/site-packages/torch/nn/modules/linear.py:114\u001b[0m, in \u001b[0;36mLinear.forward\u001b[0;34m(self, input)\u001b[0m\n\u001b[1;32m 113\u001b[0m \u001b[38;5;28;01mdef\u001b[39;00m \u001b[38;5;21mforward\u001b[39m(\u001b[38;5;28mself\u001b[39m, \u001b[38;5;28minput\u001b[39m: Tensor) \u001b[38;5;241m-\u001b[39m\u001b[38;5;241m>\u001b[39m Tensor:\n\u001b[0;32m--> 114\u001b[0m \u001b[38;5;28;01mreturn\u001b[39;00m \u001b[43mF\u001b[49m\u001b[38;5;241;43m.\u001b[39;49m\u001b[43mlinear\u001b[49m\u001b[43m(\u001b[49m\u001b[38;5;28;43minput\u001b[39;49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[38;5;28;43mself\u001b[39;49m\u001b[38;5;241;43m.\u001b[39;49m\u001b[43mweight\u001b[49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[38;5;28;43mself\u001b[39;49m\u001b[38;5;241;43m.\u001b[39;49m\u001b[43mbias\u001b[49m\u001b[43m)\u001b[49m\n",
"\u001b[0;31mTypeError\u001b[0m: linear(): argument 'input' (position 1) must be Tensor, not Dropout"
2023-12-29 11:06:25 +01:00
]
}
],
"source": [
"# Listen, um Verluste zu speichern\n",
"train_losses = []\n",
"valid_losses = []\n",
"\n",
"for epoch in range(num_epochs):\n",
" model.train()\n",
" train_loss = 0.0\n",
" for i, (inputs, labels) in enumerate(train_loader):\n",
" optimizer.zero_grad()\n",
" outputs = model(inputs)\n",
" loss = criterion(outputs, labels)\n",
" loss.backward()\n",
" optimizer.step()\n",
" train_loss += loss.item()\n",
"\n",
" # Durchschnittlicher Trainingsverlust\n",
" train_loss /= len(train_loader)\n",
" train_losses.append(train_loss)\n",
"\n",
" # Validierungsverlust\n",
" model.eval()\n",
" valid_loss = 0.0\n",
" with torch.no_grad():\n",
" for inputs, labels in valid_loader:\n",
" outputs = model(inputs)\n",
" loss = criterion(outputs, labels)\n",
" valid_loss += loss.item()\n",
"\n",
" # Durchschnittlicher Validierungsverlust\n",
" valid_loss /= len(valid_loader)\n",
" valid_losses.append(valid_loss)\n",
"\n",
" print(f'Epoch [{epoch+1}/{num_epochs}], Trainingsverlust: {train_loss:.4f}, Validierungsverlust: {valid_loss:.4f}')"
]
},
{
"cell_type": "code",
2024-01-04 11:32:45 +01:00
"execution_count": null,
2023-12-29 11:06:25 +01:00
"id": "baf1caa8-d3d9-48e8-9339-81194521528d",
"metadata": {
"tags": []
},
2024-01-04 11:32:45 +01:00
"outputs": [],
2023-12-29 11:06:25 +01:00
"source": [
"import matplotlib.pyplot as plt\n",
"\n",
"plt.plot(train_losses, label='Trainingsverlust')\n",
"plt.plot(valid_losses, label='Validierungsverlust')\n",
"plt.xlabel('Epochen')\n",
"plt.ylabel('Verlust')\n",
"plt.title('Trainings- und Validierungsverlust über die Zeit')\n",
"plt.legend()\n",
"plt.show()\n"
]
},
{
"cell_type": "code",
2024-01-04 11:32:45 +01:00
"execution_count": null,
2023-12-18 15:45:05 +01:00
"id": "8e339354-a7cc-4e8a-9323-4be41ef62117",
"metadata": {},
"outputs": [],
"source": [
"# Laden der 'kirp' Liste aus der Pickle-Datei\n",
"with open('rick.pickle', 'rb') as f:\n",
" rick = pickle.load(f)\n"
]
},
2023-12-29 11:06:25 +01:00
{
"cell_type": "markdown",
"id": "be10a487-728e-4953-a081-9103d485378c",
"metadata": {},
"source": [
"## Hauptkomponentenanalyse (PCA)"
]
},
2023-12-18 15:45:05 +01:00
{
"cell_type": "code",
2023-12-29 11:06:25 +01:00
"execution_count": 5,
2023-12-18 15:45:05 +01:00
"id": "088db0b3-8c33-41ff-a543-1b1e50c5e589",
"metadata": {
"tags": []
},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
2023-12-29 11:06:25 +01:00
"Transformierte Daten: [[-6.02552113e+01 4.57642675e+01 1.11957079e+02 ... 2.58331825e+00\n",
" 9.99342571e-01 -2.77477317e-01]\n",
" [-1.64705386e+01 9.03712725e+00 1.04837673e+01 ... 4.06859167e+00\n",
" 2.01083350e+00 1.49404086e+00]\n",
" [ 7.52348753e+00 -1.55853934e+01 -4.76301782e+01 ... -7.87604764e+00\n",
" -7.56801224e-02 8.37028680e+00]\n",
2023-12-18 15:45:05 +01:00
" ...\n",
2023-12-29 11:06:25 +01:00
" [-2.72012678e+01 4.44526098e+00 2.60063820e+01 ... 3.08321694e-01\n",
" 2.28939485e+00 -7.14920382e+00]\n",
" [-3.48027066e+01 2.27021639e+01 5.51486742e+01 ... -1.77955416e+01\n",
" 6.24722406e+00 2.32101665e+01]\n",
" [-3.98223613e+01 1.88534866e+01 5.32794498e+01 ... -1.45806809e+00\n",
" 1.18270903e+01 -2.84291311e+00]]\n",
"Varianz erklärt durch jede Komponente: [0.15056597 0.0997506 0.06070173 0.03658789 0.03530275 0.0263503\n",
" 0.02322747 0.01705354 0.01534278 0.01281486 0.01116959 0.0107472\n",
" 0.00989894 0.00906208 0.00871621 0.00813403 0.0074718 0.00708769\n",
" 0.00667045 0.00633275 0.00579241 0.00556758 0.00532382 0.00519289\n",
" 0.00476404 0.00472014 0.00457837 0.00414668 0.00399478 0.00380604\n",
" 0.00362433 0.00349278 0.00336446 0.00323228 0.00310834 0.00300595\n",
" 0.00297408 0.00285178 0.00280688 0.00273987 0.00268256 0.00263102\n",
" 0.00250513 0.00248987 0.0024505 0.0023979 0.00235971 0.00218554\n",
" 0.00217143 0.00212775 0.00210793 0.00205678 0.00202224 0.00200579\n",
" 0.00194754 0.00189606 0.00187714 0.00184969 0.00180133 0.00178537\n",
" 0.00176576 0.00172542 0.00168211 0.00167483 0.00162565 0.00159444\n",
" 0.00158667 0.00155982 0.00155534 0.00151929 0.00149558 0.00147549\n",
" 0.00146982 0.00146262 0.00143338 0.00142085 0.00140628 0.00139744\n",
" 0.00136563 0.00136169 0.00134972 0.00132027 0.00129168 0.00127963\n",
" 0.00126629 0.0012562 0.00123608 0.00122899 0.0012035 0.0011899\n",
" 0.00118094 0.00117162 0.00116552 0.00114295 0.00112631 0.00111896\n",
" 0.00110193 0.00109004 0.00108523 0.00106574 0.00106381 0.001051\n",
" 0.00104179 0.00103669 0.00103248 0.00101669 0.00100527 0.00099315\n",
" 0.00097478 0.00096486 0.00096244 0.00094792 0.00094463 0.00093107\n",
" 0.00092485 0.00090851 0.00089848 0.00089134 0.00087855 0.00087068\n",
" 0.00086397 0.00085563 0.00084342 0.00083406 0.00083064 0.00081791\n",
" 0.00080368 0.00080183 0.00079167 0.00079072 0.00078868 0.00078028\n",
" 0.00077115 0.00076662 0.00076043 0.00075196 0.0007447 0.0007332\n",
" 0.0007252 0.00072345 0.00071902 0.00070594 0.00070125 0.00069603\n",
" 0.00069029 0.00068619 0.00068012 0.00067224 0.00066615 0.00066017]\n"
2023-12-18 15:45:05 +01:00
]
}
],
"source": [
"import numpy as np\n",
"from sklearn.decomposition import PCA\n",
"from sklearn.preprocessing import StandardScaler\n",
"\n",
"# Angenommen, X ist Ihr Datensatz\n",
"# X = ...\n",
2023-12-29 11:06:25 +01:00
"X = rick\n",
2023-12-18 15:45:05 +01:00
"\n",
"# Standardisieren der Daten\n",
"scaler = StandardScaler()\n",
"X_scaled = scaler.fit_transform(X)\n",
"\n",
"# Erstellen des PCA-Objekts\n",
2023-12-29 11:06:25 +01:00
"pca = PCA(n_components=150) # Angenommen, Sie möchten 150 Hauptkomponenten behalten\n",
2023-12-18 15:45:05 +01:00
"\n",
"# Durchführen der PCA\n",
"X_pca = pca.fit_transform(X_scaled)\n",
"\n",
"# Die resultierenden Hauptkomponenten\n",
"print(\"Transformierte Daten:\", X_pca)\n",
"\n",
"# Variance Ratio für jede Komponente\n",
"print(\"Varianz erklärt durch jede Komponente:\", pca.explained_variance_ratio_)\n"
]
},
{
"cell_type": "code",
"execution_count": null,
"id": "b11bbe20-0494-4e7a-83ff-3cb0bfa82f3b",
"metadata": {},
"outputs": [],
"source": []
}
],
"metadata": {
"kernelspec": {
"display_name": "Python 3 (ipykernel)",
"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": "3.10.10"
}
},
"nbformat": 4,
"nbformat_minor": 5
}
