In a paper titled Learning biological networks: from modules to dynamics, Richard Bonneau explains why network inference is tractable in biological systems, in spite of the combinatorial nature of the problem.
- Biological networks are neither random nor designed by a known process, and therefore have yet-to-be-determined design principles. Nature does provide several clues, however, via considerations of evolution.
- Biological systems are inherently modular [...] and taking advantage of modularity is key to success in learning biological networks from data.
- Biological systems are robust and often have reproducible responses to their environment that enable replicate measurement.
- There is a lot known about the likely layout of biological networks. Several network motifs are found to be over-represented in the best characterized regulatory networks. We also know that regulatory networks are likely to be sparse (for example, most transcription factors don’t regulate most genes).
- Time-lagged correlation metrics can be used to discover regulatory relationships from microarray data.
Milo, R. et al. Network Motifs: Simple Building Blocks of Complex Networks. Science 298, 824–827 (2002). (from Uri Alon's group)
Flaherty, P., Jordan, M.I. & Arkin, A. Robust design of biological experiments. Proc. Neural Inf. Process. Symp. 18, 363–370 (2005).
Fisher, R.A. Statistical Methods, Experimental Design and Scientific Inference (Oxford University Press, Oxford, 1935).