Initial commit: AI-powered oncology insights platform

.gitignore

@@ -0,0 +1,7 @@
+.venv/
+__pycache__/
+*.pyc
+.env
+.env.*
+.DS_Store
+.ipynb_checkpoints/

.huggingface.yaml

@@ -0,0 +1,4 @@
+sdk: streamlit
+app_file: dashboard/app.py
+python_version: 3.9
+requirements_file: docs/requirements.txt

README.md

@@ -0,0 +1,51 @@
+# AI-Powered Oncology Insights Platform
+
+## Overview
+This project demonstrates how AI and real-world data (RWD) can be leveraged to extract actionable insights in oncology. It simulates a workflow for clinical decision support, drug development, and patient outcome prediction using synthetic data and modern AI tools.
+
+## Features
+- **Synthetic Oncology Data**: Simulated EHR, genomics, and treatment datasets.
+- **NLP Pipelines**: Extract cancer staging, treatments, and adverse events from unstructured clinical notes.
+- **Predictive Modeling**: Forecast patient survival or treatment response using XGBoost/LightGBM.
+- **Clustering**: Identify patient subgroups with similar biomarker or treatment profiles.
+- **Interactive Dashboards**: Visualize trends and model outputs with Streamlit or Dash.
+
+## Directory Structure
+```
+📦 oncology-ai-insights/
+├── README.md                # Project overview, goals, and demo instructions
+├── data/                    # Synthetic or anonymized RWD samples
+├── notebooks/               # Jupyter notebooks for analysis and modeling
+├── src/                     # Python scripts for preprocessing, modeling, and visualization
+├── models/                  # Saved model files and evaluation metrics
+├── dashboard/               # Streamlit or Dash app code
+└── docs/                    # Documentation and references
+```
+
+## Getting Started
+1. Clone the repository.
+2. Install requirements (see `docs/requirements.txt` or notebook cells).
+3. Explore the notebooks for data analysis and modeling.
+4. Run the dashboard app for interactive visualizations.
+
+## AI Tools & Techniques
+| Tool/Technique        | Purpose                                                                 |
+|----------------------|-------------------------------------------------------------------------|
+| NLP                  | Extract clinical features from text                                      |
+| XGBoost/LightGBM     | Predict outcomes                                                        |
+| K-Means/DBSCAN       | Patient subgroup discovery                                              |
+| Plotly/Streamlit     | Data visualization and dashboards                                       |
+| Faker/Synthea        | Synthetic data generation                                               |
+
+## Impact
+This project illustrates how AI can:
+- Support oncologists in making data-driven decisions
+- Identify clinical trial candidates
+- Uncover hidden patterns in oncology RWD
+
+## Credits
+Inspired by real-world oncology data science and Ontada-style data models.
+
+---
+
+*For demo purposes only. No real patient data is used.*

dashboard/app.py

@@ -0,0 +1,42 @@
+import streamlit as st
+import pandas as pd
+import plotly.express as px
+
+st.set_page_config(page_title="Oncology AI Insights Dashboard", layout="wide")
+st.title("Oncology AI Insights Dashboard")
+
+st.markdown("""
+This dashboard visualizes synthetic oncology patient data and demonstrates how AI can support clinical insights.
+""")
+
+@st.cache_data
+def load_data():
+    return pd.read_csv("../data/synthetic_oncology_patients.csv")
+
+df = load_data()
+
+# Sidebar filters
+cancer_types = df['cancer_type'].unique().tolist()
+st.sidebar.header("Filter Data")
+selected_cancer = st.sidebar.multiselect("Cancer Type", cancer_types, default=cancer_types)
+
+filtered_df = df[df['cancer_type'].isin(selected_cancer)]
+
+# Main dashboard
+st.subheader("Patient Demographics")
+st.dataframe(filtered_df[['patient_id', 'age', 'gender', 'cancer_type', 'stage']])
+
+st.subheader("Cancer Type Distribution")
+fig1 = px.histogram(filtered_df, x="cancer_type", color="gender", barmode="group", title="Cancer Type by Gender")
+st.plotly_chart(fig1, use_container_width=True)
+
+st.subheader("Biomarker Status vs. Treatment")
+fig2 = px.histogram(filtered_df, x="biomarker_status", color="treatment", barmode="group", title="Biomarker Status by Treatment")
+st.plotly_chart(fig2, use_container_width=True)
+
+st.subheader("Survival Time Distribution")
+fig3 = px.histogram(filtered_df, x="survival_months", nbins=8, title="Distribution of Survival Time")
+st.plotly_chart(fig3, use_container_width=True)
+
+st.markdown("---")
+st.markdown("*Demo only. No real patient data is used.*")

data/synthetic_oncology_patients.csv

@@ -0,0 +1,11 @@
+patient_id,age,gender,cancer_type,stage,diagnosis_date,biomarker_status,treatment,adverse_event,survival_months
+P001,67,F,Breast,II,2021-03-15,HER2+,Trastuzumab,None,28
+P002,59,M,Lung,III,2020-11-02,EGFR+,Osimertinib,Rash,18
+P003,72,F,Colorectal,IV,2019-07-21,BRAF-,FOLFOX,Neuropathy,12
+P004,50,M,Prostate,II,2022-01-10,AR+,Abiraterone,None,30
+P005,64,F,Ovarian,III,2021-06-18,BRCA1+,Carboplatin,Neutropenia,22
+P006,55,M,Lung,II,2022-04-05,ALK+,Alectinib,None,16
+P007,70,F,Breast,IV,2020-09-30,HER2-,Paclitaxel,Neuropathy,10
+P008,62,M,Colorectal,III,2021-12-12,BRAF+,FOLFIRI,Diarrhea,20
+P009,48,F,Ovarian,II,2022-05-25,BRCA2-,Paclitaxel,None,26
+P010,75,M,Prostate,IV,2019-02-14,AR-,Docetaxel,Fatigue,8

docs/requirements.txt

@@ -0,0 +1,3 @@
+streamlit
+pandas
+plotly

notebooks/01_data_exploration.ipynb

@@ -0,0 +1,123 @@
+{
+ "cells": [
+  {
+   "cell_type": "markdown",
+   "id": "9f06e834",
+   "metadata": {},
+   "source": [
+    "# Oncology RWD: Data Exploration\n",
+    "This notebook explores synthetic oncology patient data, simulating EHR and biomarker records."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "a51fb3f6",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "# Import libraries\n",
+    "import pandas as pd\n",
+    "import matplotlib.pyplot as plt\n",
+    "import seaborn as sns"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "48479ace",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "# Load synthetic data\n",
+    "df = pd.read_csv('../data/synthetic_oncology_patients.csv')\n",
+    "df.head()"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "ba0324c5",
+   "metadata": {},
+   "source": [
+    "## Patient Demographics"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "525868d7",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "# Demographic summary\n",
+    "df[['age', 'gender', 'cancer_type', 'stage']].describe(include='all')"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "f872f204",
+   "metadata": {},
+   "source": [
+    "## Cancer Type Distribution"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "8b89480d",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "sns.countplot(data=df, x='cancer_type', hue='gender')\n",
+    "plt.title('Cancer Type by Gender')\n",
+    "plt.show()"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "d158309e",
+   "metadata": {},
+   "source": [
+    "## Biomarker Status vs. Treatment"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "a878a045",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "pd.crosstab(df['biomarker_status'], df['treatment'])"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "50dd531c",
+   "metadata": {},
+   "source": [
+    "## Survival Analysis (Simple)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "8a9cf12f",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "sns.histplot(df['survival_months'], bins=8, kde=True)\n",
+    "plt.xlabel('Survival (months)')\n",
+    "plt.title('Distribution of Survival Time')\n",
+    "plt.show()"
+   ]
+  }
+ ],
+ "metadata": {
+  "language_info": {
+   "name": "python"
+  }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 5
+}