Thursday, June 17, 2010
Engineers have frequently used random mutations as a means of optimizing a protein or genetically engineered network. However, I don't think this is an effective optimization or design process. Planned mutations, such as the one used by the immune system, rely on homologous recombination rather than random mutations. Using recombination, we can plan the mutation events, therefore perform a much more predictable optimization. The optimization process can even be simulated computationally.
Thursday, June 3, 2010
Suppose Amazon and eBay want some algorithm that finds common patterns is the customer purchasing history. Suppose a third company is developing exactly such an algorithm. It is cost-effective for Amazon and eBay to outsource this pattern finding problem to this third company, because the third company only charges half of what is required to get the job done. The third company is able to charge half the normal price because it gets payment from both companies. This can be considered evolution of modularity in economics, or niche finding.
The same reasoning might apply to modular structures in biology. It is perhaps easiest to think of an ecosystem first, where each species is like a company. If one species provides a function that is needed by the other species, then the ecosystem depends on that first species. In other words, the ecosystem will probably evolve so that there is a reserved seat for the first species. Now, moving the analogy to population of cells or population of genes within cells is a bit different, but I think some of the analogies still apply. The underlying idea is that the system favors those species that are necessary for the whole system to survive. And the reduction of cost is the incentive for new species to evolve and take a specific role in the system.