Artificial Neural Network made from DNA

A piece of DNA

The researchers at California Institute of Technology (Caltech) have created an artificial neural network out of DNA which, on asking questions, gave the correct answers.

The researchers at Caltech under the leadership of Lulu Qian have created a circuit of interacting DNA molecules that can recall memories based on incomplete patterns , just as brain can. The researchers based their biochemical neural network on a simple model of neuron, called a linear threshold function. The model neuron receives input signals, multiplies each by a positive and a negative weight and if the weighed sum of inputs surpass a certain threshold , the neuron fires, producing an output.The Caltech used molecules because they knew that before the neural-based brain developed, single-celled organisms showed limited forms of intelligence which was the result of interaction of molecules that spurred it to search for food and avoid toxins. Thus, these molecules acted like circuits processing and transmitting information and computing data.

The Artificial Neural Network

They used DNA molecules specifically because these molecules interact in specific ways determined by the sequence of their four bases-adenine(abbreviated as A), cytosine(C), thymine(T), and guanine(G). The scientists can encode these sequences into strands of DNA molecules, and programmed them to function in a predefined way. So, the researchers created 4 highly simplified artificial neurons comprising of 112 strands of DNA, and each strand programmed differently. After this, the researchers played a mind-reading game with the neural network in which it tries to identify a scientist, after training it to know 4 scientists, whose identities are each represented by a specific, unique set of answers to 4 yes-or-no questions. A human player , to question the brain, placed strands of DNA that hinted at the answer, and with the provided clues the artificial brain was able to produce the correct answer. In this way, the brain could correctly answer even when it lacked enough information or any of the clues had contradictory information. The researchers played this game with the network using 27 different ways of answering the questions out of 81 total combinations, and each time, the brain responded correctly.

So, we can tell that this is a replication of intelligence. And, I think, on complex programming and construction of this neural network, we can make a brain of a lower vertebrate.

Share