Medicine and Surgery HT

Course Objectives: This exam will assess understanding of the fundamental concepts related to complex network analysis, modularity, centrality measures, and their specific applications in the field of biomedicine and precision medicine, as described in the course material. Particular emphasis will be placed on the distinction between data and information, the modeling of biological systems through networks, and their functional and clinical interpretation. Exam Sections: I. Introduction to Networks and Epistemological Foundations • Data and Information: ◦ Definition of data and information . ◦ The role of context and questions in transforming data into information . ◦ Concrete examples: phone numbers, Earth's temperature, earthquakes, drug response, London Underground . ◦ Critique of context-less artificial intelligence . ◦ The importance of "knowing how to pose problems" (Gaston Bachelard) . ◦ The case of the London Underground map: distortion of reality to extract useful information (Harry Beck) . ◦ "To know is to deform the real" (Carlo Emilio Gadda) . ◦ Cognitive metaphor as a tool for understanding (subway network as an electrical circuit) . ◦ "Seeing better does not mean seeing more" and Borges' legend of the perfect map . ◦ The "question-driven" approach versus the "data-driven" approach in biomedical science . ◦ Networks as interpretive models of the real world, not as reality in themselves . II. Networks and Modularity • Basic Network Definitions: ◦ Distinction between network and graph . ◦ Network components: nodes (vertices, V) and edges (links, E) . ◦ Types of networks: directed, undirected, weighted . ◦ Adjacency matrix representation . ◦ Structural properties: density, paths, diameter . ◦ Density calculation . ◦ Distance between nodes (shortest path), average path length, and diameter . ◦ Relevance of density and average path length in precision medicine (e.g., oncology, neurodegeneration, rare diseases) . • Bipartite Networks: ◦ Definition and characteristics (two disjoint sets of nodes) . ◦ Examples in clinical settings (patient-phenotypes) . ◦ Projection of bipartite networks (patient-patient network, phenotype-phenotype network) . ◦ Examples of bipartite networks in biology and medicine (predator-prey, disease-RNA, drug-target, gene-transcription factors, patient-anatomical locations) . • Modularity in Networks: ◦ Concept of community or module (not rigid) . ◦ Biological functions as a result of molecular group interaction . ◦ Modularity as an evolutionary principle (specialization, resilience, evolvability) . ◦ Formal definitions of communities: weak community, strong community, clique . ◦ Identification methods: connected subnetwork vs. locally dense region . ◦ The principle of "guilt by association" (homophily/assortative mixing) . ◦ Module identification: cut-based vs. density-based strategies . ◦ Cosine distance and hierarchical clustering for modularity . ◦ The modularity measure Q (Newman and Girvan): definition, calculation, interpretation, null model (configuration model), and resolution parameter (γ) . ◦ The Louvain algorithm: operation (local optimization and regrouping phases), efficiency, limitations . ◦ Patient stratification through network analysis (e.g., cardiac risk in cancer patients) . III. The Importance of Being "Central" • Important Nodes and Centrality Measures: ◦ Concept of central node and "the three Cs" . ◦ Degree centrality and "hub" nodes . ◦ Scale-free networks and degree distribution . ◦ Connector nodes (bow tie motifs) and their role . ◦ Node roles in modular networks (central, peripheral, connector nodes) . ◦ Closeness centrality: definition and interpretation . ◦ Betweenness centrality: definition and interpretation . ◦ Eigenvector centrality: definition and interpretation (derived popularity) . • Applications and Caveats: ◦ Correlation between degree and protein essentiality (Gerstein experiment) and its criticisms . ◦ Centrality and cancer (Vidal, Sun and Zhao, oncogenic hubs) . ◦ Standard network analysis protocol: construction, module search, enrichment, important node identification . IV. Network Architectures and Their Evolution • Random Networks: ◦ The Erdős-Rényi model G(n,p) . ◦ Percolation transition and the giant component . ◦ Limitations of random networks in describing real systems . • Heavy-Tailed and Scale-Free Networks: ◦ Power-law degree distribution . ◦ Robustness to random attacks and vulnerability to targeted attacks (hubs) . ◦ Power laws and phase transitions in complex systems . • Network Evolution: ◦ Growth and preferential attachment (Barabási-Albert model) . ◦ Criticisms of the Barabási-Albert model . ◦ General criticisms of scale-free networks (Evelyn Fox Keller, statistical limitations, metaphorical nature of network properties) . V. Classification of Molecular Networks and Disease Modules • Physical Interaction Networks: ◦ Interactome: definition and types of interactions (protein-protein, DNA-protein, protein-RNA, miRNA-mRNA) . ◦ Stable vs. transient interactions and "just-in-time assembly" . ◦ Experimental methods for PPI: Yeast Two-Hybrid (Y2H) and Affinity Purification/Mass Spectrometry (AP/MS), their advantages and limitations . ◦ Impact of mutations on genotype-phenotype (nonsense, missense, "edgetic" effects) . ◦ Alternative splicing and protein isoforms . ◦ Static vs. context-specific interactome . ◦ PPI Databases (HPRD, STRING, OncoPPi, HuRI) . ◦ Integration of gene expression data with PPI networks . • DNA-Protein Interaction Networks (Transcriptional Regulatory Networks): ◦ Representation as directed and bipartite networks . ◦ ChIP-sequencing (ChIP-Seq) method . ◦ Databases (TRANSFAC, TRRUST) . • Association Networks: ◦ Definition (nodes as genes, proteins, metabolites, phenotypes; edges as statistical/functional associations) . ◦ Applications in precision medicine (functional modules, guilt-by-association, patient stratification, drug prediction) . • Gene Co-expression Networks: ◦ Definition (genes with similar expression profiles) and synexpression groups . ◦ Mechanisms of co-expression (operons, epigenetics, miRNAs, shared cell signaling) . ◦ Similarity measure: Pearson correlation (Rij), alternatives, multiple testing correction . ◦ Choosing the significance threshold (hard vs. soft thresholding, giant component) . ◦ Weighted vs. unweighted, signed vs. unsigned networks . ◦ Co-expression modules: identification (sentinel genes, hierarchical clustering, k-means, Louvain) and characterization (eigengene, correlation with clinical traits) . ◦ Omics integration (co-expression with PPI) . ◦ Differential co-expression networks: distinction from differential expression, methodologies (simple correlation difference, Fisher's z-transformation), clinical applications (Alzheimer's, cancer, autoimmune diseases) . • Disease Modules: ◦ The Harvard Hypothesis (HNM): disease as dysfunction of molecular circuits (disease module) . ◦ Identification of "seed genes" and associated databases (COSMIC, ClinVar, OMIM, TCGA) . ◦ Limitations of the interactome (incompleteness, noise, non-specificity) . ◦ Concept of module malfunction (loss/gain of function, interaction alterations) . ◦ Criticisms and validation of HNM (Choobdar, Agrawal studies) . ◦ Approaches (molecular, non-reductionist, personalized) . ◦ Genotype-phenotype relationship mediated by the module . ◦ Distance between disease modules and correlation with functional/clinical/comorbidity similarity . ◦ Methods for identifying disease genes (direct, indirect) . ◦ Algorithms for disease module prediction: vicinity-based, connectivity-based (DIAMOnD), diffusion-based (Random Walk with Restart - RWR) . • Drug Repurposing: ◦ Definition and advantages . ◦ Network-based methods for repurposing: modeling drug-gene-protein-disease relationships . ◦ Measures of topological proximity between drug and disease: kernel distance (dk), center distance (dcc), separation distance (dss), shortest path distance, direct connect, RWR . ◦ Clinical example of repurposing (soluble guanylate cyclase activators for ischemic stroke)

Lecture notes provided by the professor