ŷ������

“The substrates within each CDR that are frequently seen mutated are defined as “hotspots��. They are described by preferences for purines, rather than pyrimidines, as well as for particular codons, or codon motifs within the sequence. The fact that mutation in a hotspot can create or delete other hotspots indicates a higher order structure to the mutation process than that which is currently observable.
McKay Brown, Mary Stenzel-Poore, Susan Stevens, Sophia K. Kondoleon, James Ng, Hans Peter Bachinger, and Marvin B. Rittenberg. Immunologic memory to phosphocholine keyhole limpet hemocyanin. Journal of Immunology, 148(2):339��346, January 1992.”
― Laura F. Landweber, Evolution as Computation

“These potential advantages of DNA computing over the traditional approach and the seminal experimental work of Adleman, demonstrating the practical in vitro implementation of a DNA algorithm for solving an instance of the Hamiltonian path problem, caused a strong increase of interest in DNA computing over the past years. Although the set of “bio-operations�� that can be executed on DNA strands in a laboratory (including operators such as synthesizing, mixing, annealing, melting, amplifying, separating, extracting, cutting, and ligating DNA strands) seems fundamentally different from traditional programming languages, theoretical work on the computational power of various models of DNA computing demonstrates that certain subsets of these operators are computationally complete. In other words, everything that is Turing-computable can also be computed by these DNA models of computation. Furthermore, it has also been shown that universal systems exist, so that the programmable DNA computer is theoretically possible.
The algorithms for DNA computing that have been presented in the literature use an approach that will not work for NP-complete problems of realistic size, because these algorithms are all based on extracting an existing solution from a sufficiently large initial population of solutions. Although a huge number (�� 1012) of DNA molecules (i.e., potential solutions to a given problem) can be manipulated in parallel, this so-called filtering approach (i.e., generate and test) quickly becomes infeasible as problem sizes grow (e.g., a 500-node instance of the traveling salesman problem has > 101000 potential solutions).”
― Laura F. Landweber, Evolution as Computation

“It is now plausible at the molecular level to conceive of concerted, non-random changes in the genome guided by cellular computing networks during episodes of evolutionary change. Thus, just as the genome has come to be seen as a highly sophisticated information storage system, its evolution has become a matter of highly sophisticated information processing.”
― Laura F. Landweber, Evolution as Computation