A DNA computer has been developed that can play tic-tac-toe against a human and never lose. The device uses a complex mixture of DNA strands and DNA-based enzymes to determine where it should place its nought or cross.
The computer, dubbed MAYA, was designed by Milan Stojanovic at Columbia University in New York and Darko Stefanovic from the University of New Mexico in Albuquerque. In other DNA computers, all of the elements are mixed in a single test tube and the answer to a calculation is deduced from the product of the reaction. MAYA represents a significant advance because it is the first interactive system. It consists of nine tiny wells arranged in a three-by-three grid, as in the game. Each of the wells contains a different mixture of DNA strands and enzymes, which act like the logic gates found in a conventional computer.
If the enzymes in a well are activated, they cleave the DNA strands, which then produce green light 鈥� this is how MAYA signals a move (see Graphic). The human player moves by adding a single strand of DNA with a sequence that 鈥渃odes鈥� for the square where they want to play to all of the wells. These DNA molecules activate particular enzymes. Some are activated by just one strand, while others need two or three inputs, each representing a previous move.
Advertisement
The enzymes are carefully constructed and distributed so that, after the human player鈥檚 move, the DNA-cleaving enzymes are activated only in one well. This signals to the computer that it can then take its move. And because tic-tac-toe is so simple, the computer can be designed so that it always achieves a win or a draw. For example, Stojanovic has lost to MAYA more than a hundred times. The design of the computer will be published in a future issue of Nature Biotechnology.
鈥淚t鈥檚 lovely work,鈥� says Peter Bentley, a computer scientist linked to University College London. But he warns that a system that can never be extended beyond playing tic-tac-toe will always remain a novelty.
More complex computational tasks could perhaps be tackled with different arrangements of the logic gates. But the researchers agree that the approach is unlikely ever to rival silicon computers, because people have to interfere to operate the gates and each gate can only be used once 鈥� the enzymes cannot be reset. Instead Stojanovic and Stefanovic hope to develop simple decision-making molecules that can operate within the body. These could assess genetic faults in a living cell, for example, and then either kill or repair it.