Introduction to Machine Learning for analysis of biological data with R Présentiel

Day 1: R Essentials & Data Quality
Implement development best practices and prepare a robust dataset

• Morning: Environment and Syntax
  • Project Management: Directory structuring, reproducibility best practices, and version control
  • HPC Environment: Introduction to High-Performance Computing (HPC) for executing R scripts at scale
  • R refresher : Review of fundamental syntax and data structure manipulation (dataframes, plot)

• Afternoon: Data Preparation
  • Quality Control (QC): Outlier identification, missing data handling, and distribution assessment
  • Data Scaling: Normalization techniques, data transformation, and centering/scaling to prepare data for statistical models

Day 2: Dimensionality Reduction

Extract relevant signals and visualize high-dimensional datasets

• Morning: Principal Component Analysis (PCA):
  • PCA principle, implementation in R, and component interpretation (explained variance, biplots)
  • Hands-on practice

• Afternoon: PCA and non linear approach
  • PCA (Part 2): Advanced applications and hands-on practice
  • Introduction to UMAP (highly suited for omics data)

Day 3: Clustering & Linear Regression

Identify underlying structures (expression profiles) and introduce modeling

• Morning: Unsupervised Learning (Clustering)
  • K-means: Algorithm principles, determining the optimal number of clusters (elbow method, silhouette score)
  • Hierarchical Clustering: Distance metrics, linkage methods, and dendrogram interpretation

• Afternoon: Supervised Modeling
  • Clustering (Part 2): Advanced applications and hands-on practice
  • Linear Regression: Model building, assumption testing (residuals, homoscedasticity), and coefficient interpretation

Day 4: Advanced Modeling (Half-Day)

Expand statistical foundations into generalized models.

• Morning: Advanced Applications & Wrap-up
  • Advanced Regression Techniques: introduction to Generalized Linear Models (logistic regression, count models (Poisson / negative binomial))
  • Wrap-up: Q&A session and course review

Machine learning and multivariate statistics have become central to the analysis of biological and clinical data. This hands-on course offers a practical introduction to the complete analysis pipeline in R, from reproducible project setup and quality control to dimensionality reduction (PCA, UMAP), clustering (k-means, hierarchical), and regression modelling (linear and generalized linear models). The emphasis is on practical implementation and interpretation rather than mathematical formalism: participants learn to apply each method to real datasets, tune its parameters, and critically interpret the results. Examples are drawn from transcriptomics, clinical and ecological data to build transferable skills.

• Set up a reproducible R project and run analyses in an HPC environment
• Perform quality control, handle missing data, and normalize/scale biological datasets
• Apply and interpret dimensionality reduction methods (PCA, UMAP), including parameter tuning
• Implement unsupervised clustering (k-means, hierarchical) and determine the optimal number of clusters
• Build, diagnose, and interpret regression models (linear and generalized linear models)

Ressources pédagogiques: files in PDF format will be made available to the trainee.

Equipements provided: digital resources ; to be brought by students: laptop.