luokai0 Computational Biology Expert
Before Starting
- Molecular, cellular, or systems level?
- Simulation, prediction, or data analysis?
- Which software ecosystem? (Python, R, GROMACS, NAMD)
Core Expertise Areas
Molecular Dynamics
Force fields: AMBER, CHARMM, GROMOS define atom interactions. Integration: Verlet or leapfrog algorithm integrates equations of motion. Periodic boundary conditions: avoid surface effects, simulate bulk behavior. Thermostat and barostat: maintain temperature and pressure in NPT ensemble. Free energy: umbrella sampling, metadynamics, alchemical perturbation methods.
Protein Structure Prediction
AlphaFold2: deep learning, near-experimental accuracy for single domains. ESMFold: language model based, faster but slightly less accurate than AlphaFold2. Homology modeling: template-based for sequences with known homologs. Intrinsically disordered: not well-handled by structure prediction, need ensemble. Structure validation: DOPE score, MolProbity, Ramachandran plot analysis.
Systems Biology
ODE models: deterministic kinetic models of reaction networks. Stochastic simulation: Gillespie algorithm for small molecule numbers. Boolean networks: coarse-grained on/off gene regulation models. Metabolic flux analysis: FBA predicts steady-state fluxes in metabolic networks. Parameter estimation: MCMC and optimization for fitting models to data.
Network Biology
PPI networks: protein-protein interaction networks, hub proteins. Gene regulatory networks: transcription factor target relationships. Network topology: degree distribution, clustering coefficient, shortest paths. Community detection: modules in biological networks correspond to functional units.
Best Practices
- Always validate simulation against experimental observables
- Use sufficient simulation time to ensure convergence
- Apply appropriate statistical tests to simulation data
- Document all parameters and software versions for reproducibility
Common Pitfalls
| Pitfall | Fix |
|---|---|
| Insufficient simulation length | Check convergence with multiple metrics |
| Wrong protonation states | Calculate pKa and set protonation at simulation pH |
| Overinterpreting AlphaFold structures | High confidence regions are reliable, low confidence are not |
| Ignoring model identifiability | Many parameters may give same fit to data |
Related Skills
- molecular-biology-expert
- bioinformatics-expert
- machine-learning-expert