Scientists from the California Institute of Technology have created an artificial neural network (or a tiny brain, in the words of the lead scientist) from DNA strands that interact with biochemical inputs.

The artificial neurons of this network can take incomplete inputs, interact with each other, and come up with a complete conclusion. This is what the human brain does on a much more complex scale. It's also a principle scientists have used for computing and robotics.

The building block of the Caltech neural network is double-stranded DNA molecules with loose ends. These loose ends then receive the input of single-stranded DNA that binds with the double-stranded DNA, which, through DNA strand replacement, releases an output DNA strand from the double-stranded DNA.

Using this input-output mechanism, the Caltech team assembled four neurons that give out specific DNA strand outputs that serve as both 'yes' or 'no' indicators in themselves and also inputs strands into other neurons.

The end result is a neural network capable of churning out outputs for all four neurons in the absence of input for all four neurons.

The specific experiment the Caltech researchers conducted was guessing the identity of four scientiststs. The four neurons corresponded to the following four questions.

Then, by inputting only 'yes' for Q3 and 'no' for Q4, for example, the neural network is able to deduce 0 1 1 0, or Rosalind Franklin. The research claims that there are 27 combinations of incomplete inputs that would yield a correct identification.

Moreover, if the incomplete information doesn't describe anyone (for example 'yes' for 2 and 'no' for 3), the neural network would return both 'yes' and 'no' for all four neurons, indicating an input error.

Biochemical systems with artificial intelligence-or at least some basic, decision-making capabilities-could have powerful applications in medicine, chemistry, and biological research, stated Caltech in a press release.

Lulu Qian, a Caltech senior postdoctoral scholar in bioengineering, is the lead author of the study, which is publsihed in the July 21 issue of journal Nature. Below are YouTube videos from her that explains this study. The three images used in this article are from the YouTube videos.