Components of a living organism, from organs and tissues to single cells and subcellular compartments, exchange and process numerous molecular signals in order to coordinate their activity. When these components fail, they generate characteristic signals that often trigger self-repair processes but can also cause disease when left unchecked. Engineered biomolecular systems that complement natural defense and repair mechanisms could lead to novel diagnostic and therapeutic tools. We call these systems "biological automata" because they resemble electromechanical automata in that they sense disease-related signals and convert them to a response using biological computing modules. Biological automata might be constructed as anything from molecular networks operating inside cells to "programmed" cells and their networks operating in organisms.

Biological molecular networks that process and convert information encoded in endogenous molecular signals are comprised of a molecular computation module that interacts with the signals via a sensory module, processes them in a programmed fashion and translates the results of the processing into biological activity via an actuation module. A complex condition may often be approximated by a logical (Boolean) expression on biological signals - levels of gene expression as manifested by mRNA and protein levels, presence of mutations, and concentrations of metabolites among other things. Within this framework, detecting a complex condition corresponds to evaluating a logical expression that imposes a relation between the presence or absence of the individual signals  expression's variables  and the presence or absence of a condition.

We have constructed a de novo molecular information-processing gene network that operates in human kidney cells. This automaton is based on RNA interference (RNAi), a mechanism for RNA-guided regulation of gene expression. We show that the RNAi pathway in human cells can form a molecular computing module capable of evaluating arbitrary Boolean expressions on endogenous cues. We experimentally demonstrate in human kidney cells the direct evaluation using exemplary expressions in standard forms with up to five logic variables.