Cyclic mixture mutagenesis is based upon oligonucleotide-directed mutagenesis, a standard technique in molecular biology for introducing specific mutations into new copies of DNA strands in vitro. The technique involves performing DNA replication in the presence of many copies of a short DNA strand (3) which will bind to the long template strand being copied and end up becoming incorporated into the new copy as it is formed.
However, as traditionally applied, this technique is limited to just introducing a mutation, and nothing more. We have discovered, and verified in computater simulations (with real-world experiments to be performed soon), that when repeated cycles of DNA replication are performed in the presence of appropriately designed mixtures of different oligonucleotides, the resulting system can act as a sort of DNA "machine," in which subsequent generations of descendant strands undergo a definite sequence of programmed changes, dependent on the initial sequence, and clocked by the cycles of replication. (Thus the name "cyclic mixture mutagenesis.") The resulting machines can be viewed as computational systems with unlimited power (in principle); we estimate that we can design a particular mixture of at most a few thousand species of oligonucleotides which would suffice to execute an arbitrary computer program encoded on the initial strand. (4) Additionally, we can take advantage of the massive parallelism inherent in the system by introducing random choices into the computation. Results of such nondeterministic computations can be easily obtained by screening for strands containing a special sequence that the system is programmed to produce once a result has been found.
Moreover, our "mutagenesis machine" has potential applications in living systems: it could explain how dividing cells differentiate from each other during the development of complex multicellular organisms, and also how cell lines seem to be able to count the number of their divisions. Both of these operations could in principle be carried out by having definite concentrations of certain mutagenic oligonucleotides present in the nucleus during replication. One could also imagine designing intelligent DNA drugs consisting of modified, biostable oligonucleotides that, when injected into developing cells, cause different desired changes in chromosomal DNA to occur when cells divide.
Preliminary to any eventual in vivo applications, we have been exploring the use of our techniques to create DNA machines in vitro, using standard PCR techniques to perform DNA replication. We have designed several particular sets of oligonucleotides that are intended to implement various simple DNA machines, and we have verified, using detailed computer simulations of the physical chemistry (therodynamics and kinetics) of inter-strand DNA binding, that these machines should work as expected. We have selected the details of our experimental protocols and have already begun in vitro experiments. Let us now discuss the major results obtained to date.