Biocomputers: from test tubes to live cells

Abstract

Biocomputers are man-made biological networks whose goal is to probe and control biological hosts--cells and organisms--in which they operate. Their key design features, informed by computer science and engineering, are programmability, modularity and versatility. While still a work in progress, biocomputers will eventually enable disease diagnosis and treatment with single-cell precision, lead to "designer" cell functions for biotechnology, and bring about a new generation of biological measurement tools. This review describes the intellectual foundation of the "biocomputer" concept as well as surveys the state of the art in the field.

Figure 1

Silicon-based and biomolecular computers compared. A, General layout of a conventional computer. B, General layout of a biomolecular computer. Molecular inputs are converted into an internal molecular representation via a number of molecular sensors. Those molecules are in turn computationally processed by “molecular software” and “hardware”. Molecular hardware would normally contain some invariant mechanisms (e. g. specific regulatory pathways) that can be easily reconfigured to implement different algorithms.

Figure 2

A decision-making finite state machine. A, General layout of the system. A DNA molecule (top) encodes internal representation of the conditions to be tested, and the left-most segment is labeled with a single-stranded extrusion indicating the current state. A condition-checking “black box” that is external to the computation and constitutes input peripherals produces one of the two “transition molecules”, depending on the condition state. B, The workings of the input peripherals. An mRNA molecule whose concentration is a condition of interest physically interacts with appropriately designed transition molecules and alters their composition in concentration-dependent manner using strand exchange mechanism. C, The underlying chemical process: transition molecules generated by the input peripherals chew on the DNA molecule and move it to the next step, generating either a Yes-state or No-state encoding sticky end. The newly exposed sticky ends are substrates for the next round of “chop-off” with a pair of applicable transition molecules, and so on until the final state is reached. The sequence exposed in this final configuration triggers more downstream processing resulting in a release of drug-like molecule (not shown).

Figure 3

Biocomputers that employ normal logic forms. A, Disjunctive (DNF) and conjunctive (CNF) logic forms B, The internal structure of an RNAi-based biocomputer and the core logic function it computes. The function can be reprogrammed by adding or removing small RNA targets as well as moving the targets among the mRNA reporter constructs. The same target can appear in multiple reporter constructs as well. C, Computing logic relations between biological inputs once the input peripherals are established.