Recent work has proven DNA computing to be a powerful tool for solving very hard problems like the Hamiltonian Path (Traveling Salesman) Problem. This work has spawned a new research field, called DNA computing, to study in depth novel applications of this new medium. One recent motivator in this field has been to replace silicon based computers for certain applications with DNA-based computers. My theme is basic research to provide these DNA computers with suitable memories. I present substantial evidence that DNA memories can reliably store data in such a way that information can be retrieved associatively and semantically. The massively parallel nature of DNA lends a means of retrieval that is very fast (order of minutes) by current biotechnologies. DNA memories were prototyped and evaluated on two challenging problems in silico using EdnaCo, a virtual test tube simulator. Specifically, DNA memories successfully retrieved solutions for the problems of Recognizing Textual Entailment (RTE) by simple methods that perform competitively with far subtler lexical methods; they also distinguished genomes of biological organisms without significant prior classification effort. The capacity of DNA memories is shown to be at least as dense as that of Hopfield Memory, the standard associative memory benchmark implemented with Neural Networks. The contribution of this research is designs for DNA memories that are not only capable of solving problems hitherto very challenging to conventional computers, but are feasible to implement with conventional powerful biotechnologies, e.g. DNA chips and microarrays.
展开▼
