Systems biology Study Guide

📖 Core Concepts Systems Biology – computational & mathematical analysis of whole biological systems; emphasizes holistic over reductionist view. Emergent Property – a system‑level behavior (e.g., robustness, adaptation) that cannot be predicted by studying components in isolation. Multi‑omics – simultaneous measurement of genomics, transcriptomics, proteomics, metabolomics, etc., providing system‑wide quantitative data. Model Types – Boolean, Petri net, ODE/PDE, stochastic, Bayesian, agent‑based, rule‑based, state‑space; each abstracts the system at a different level of detail. Top‑down vs Bottom‑up – data‑driven discovery (top‑down) versus mechanistic, kinetic reconstruction (bottom‑up). Constraint‑Based Reconstruction – uses mass‑balance constraints (e.g., Flux Balance Analysis) to predict steady‑state fluxes without kinetic parameters. Iterative Cycle – theory → computational model → hypothesis → experiment → model refinement. --- 📌 Must Remember Goal: generate testable hypotheses, predict behavior, and design interventions. Top‑down advantage: genome‑wide insight, fast hypothesis generation from correlation patterns. Bottom‑up advantage: quantitative translation of in‑vitro data to in‑vivo predictions (e.g., PBPK models). ODE Mass‑action rate: $v = k\prodi [Xi]^{\nui}$ (units: concentration·time\(^{-1}\)). Flux Balance Analysis (FBA) – linear programming problem: $$\max{v} \; c^{T}v \quad \text{s.t.} \; S v = 0,\; v{\min} \le v \le v{\max}$$ where \(S\) = stoichiometric matrix, \(c\) = objective coefficients. Gillespie algorithm – exact stochastic simulation; key for low‑copy‑number species. Boolean attractor ↔ stable cellular phenotype (e.g., differentiation state). AI role: pattern detection, network inference, integrative multi‑omics modeling, virtual screening. --- 🔄 Key Processes Methodological Cycle Formulate theory → build computational model → generate hypotheses → perform experiments → validate → refine model. Model Construction (Bottom‑up) Define pathway → draw network diagram → choose kinetic law (mass‑action, Michaelis‑Menten) → write ODEs → estimate parameters (literature, fitting) → handle unknowns (sensitivity, Bayesian inference) → validate with data. Constraint‑Based Reconstruction Assemble genome‑scale network → impose mass‑balance (\(Sv = 0\)) → set bounds → define objective (e.g., biomass) → solve linear program → interpret flux distribution. Multi‑omics Integration Preprocess each omics layer → map to common identifiers → build multi‑layer network (genes ↔ proteins ↔ metabolites) → apply correlation, Bayesian, or machine‑learning methods → extract modules/hubs. --- 🔍 Key Comparisons Top‑down vs Bottom‑up Data reliance: correlation‑heavy vs mechanistic detail. Scale: genome‑wide vs pathway‑focused. Deterministic ODE vs Stochastic (Gillespie) Molecule count: high (continuous) vs low (discrete). Outcome: smooth trajectories vs probability distributions. Boolean vs Quantitative (ODE/PDE) State: binary (ON/OFF) vs continuous concentrations. Parameter need: none vs many kinetic constants. FBA vs Kinetic Modeling Requirement: only stoichiometry vs detailed rate laws. Prediction: steady‑state fluxes vs time‑course dynamics. --- ⚠️ Common Misunderstandings “All systems biology needs kinetic parameters.” – False; constraint‑based models work without them. “Top‑down models are less useful.” – They uncover novel interactions and guide hypothesis generation. “Stochastic models are always slower.” – Approximate methods (τ‑leaping) can be faster than deterministic stiff ODE solvers. “AI replaces mechanistic modeling.” – AI augments, not substitutes, mechanistic insight; results still need biological validation. --- 🧠 Mental Models / Intuition Network as a city map: hubs = major highways (critical for traffic flow), modules = neighborhoods (functional clusters). Bottom‑up = building a house brick by brick; top‑down = looking at the finished skyline to infer the layout. FBA = water flowing through pipes at steady state; ODEs = water level changing over time. --- 🚩 Exceptions & Edge Cases Hybrid models – combine ODEs for high‑abundance species with stochastic simulation for rare molecules. Delayed stochastic simulations – needed when transcription‑translation lag is comparable to reaction timescales. Partial differential equations – only when spatial gradients (e.g., morphogen patterns) matter. Non‑identifiable parameters – use ensemble modeling or fix certain parameters based on literature to avoid over‑fitting. --- 📍 When to Use Which Use Boolean models when kinetic data are scarce and the network is large (>100 nodes). Use ODEs for well‑characterized pathways with quantitative kinetic measurements. Use Stochastic (Gillespie) for low‑copy‑number species or when noise drives phenotype (e.g., gene‑expression bursts). Use FBA for genome‑scale metabolic networks where only stoichiometry is known. Use AI/ML for pattern discovery in high‑dimensional multi‑omics or for predicting drug‑target interactions. Use Petri nets when you need to capture concurrency and resource constraints explicitly. --- 👀 Patterns to Recognize Correlation clusters → potential co‑regulated modules (top‑down cue). Steady‑state flux patterns that maximize biomass → typical FBA solution. Attractors in Boolean simulations → stable cell fates or disease states. Stiff ODE systems → presence of fast and slow reactions; need implicit solvers. Hub‑and‑spoke topology → likely points for therapeutic intervention. --- 🗂️ Exam Traps “All omics data are independent.” – Many integration methods rely on shared identifiers; ignoring cross‑layer links loses information. Choosing a model because it “sounds more advanced.” – The correct choice hinges on data availability, not sophistication. Confusing mass‑action with Michaelis‑Menten kinetics – the former is proportional to reactant concentrations; the latter includes saturation terms. Assuming FBA predicts dynamics – FBA yields only steady‑state fluxes; time‑course predictions need kinetic models. Over‑relying on AI predictions without validation – AI outputs are hypotheses, not proven mechanisms. ---