Cancer Risk Prediction with Machine Learning for Bioinformatics

Day 1: Bioinformatics Data Foundations & ML Readiness

• Cancer risk prediction use-cases (screening, stratification, prognosis vs risk)
• Data types & formats: clinical tables, gene expression (RNA-seq/microarray), mutation features, methylation basics
• Python foundations for bioinformatics ML: NumPy, Pandas, Matplotlib; dataset structuring
• Data preprocessing: missing values, encoding, scaling, normalization concepts for omics
• Exploratory data analysis (EDA): distributions, class imbalance, correlations, feature sanity checks
• Feature engineering: pathway scores intro, aggregation strategies, variance filtering
• Tools: Jupyter/Colab, Pandas, NumPy, Matplotlib, Scikit-learn

Day 2: ML Models for Cancer Risk Prediction (Hands-on)

• Supervised learning models: logistic regression, random forest, XGBoost/GBM overview, SVM
• Handling imbalance: class weights, sampling methods (SMOTE concept + safe usage)
• Model validation: train/val/test split, stratification, cross-validation, leakage prevention
• Performance metrics: ROC-AUC, PR-AUC, precision/recall, F1, confusion matrix; thresholding for decision-making
• Interpretability: feature importance, coefficients, permutation importance, SHAP overview
• Case study lab: build a baseline risk classifier from bioinformatics + clinical features
• Tools: Scikit-learn, (optional) XGBoost, Seaborn/Matplotlib, SHAP (intro)

Day 3: Advanced Techniques, Reproducibility & Research-Grade Reporting

• Advanced feature selection: L1 regularization, RFECV, mutual information, stability considerations
• Hyperparameter tuning: GridSearchCV/RandomizedSearchCV, model comparison table
• Pipeline best practices: preprocessing + modeling with Pipeline, reproducible seeds, documentation
• Reporting: figures, metric tables, model cards, limitations, bias/fairness basics for healthcare ML
• Mini-project: end-to-end cancer risk prediction pipeline + final report structure
  Optional extension: survival analysis direction (what changes, what to read next)

Tools: Scikit-learn Pipelines, GridSearchCV, Jupyter/Colab

Take your research to the next level with NanoSchool.