Dynamic self-assembly in living systems as computation

Abstract

Biochemical reactions taking place in living systems that map different inputs to specific outputs are intuitively recognized as performing information processing. Conventional wisdom distinguishes such proteins, whose primary function is to transfer and process information, from proteins that perform the vast majority of the construction, maintenance, and actuation tasks of the cell (assembling and disassembling macromolecular structures, producing movement, and synthesizing and degrading molecules). In this paper, we examine the computing capabilities of biological processes in the context of the formal model of computing known as the random access machine (RAM) [Dewdney AK (1993) The New Turing Omnibus. Computer Science Press, New York], which is equivalent to a Turing machine [Minsky ML (1967) Computation: Finite and Infinite Machines. Prentice-Hall, Englewood Cliffs, NJ]. When viewed from the RAM perspective, we observe that many of these dynamic self-assembly processes – synthesis, degradation, assembly, movement – do carry out computational operations. We also show that the same computing model is applicable at other hierarchical levels of biological systems (e.g., cellular or organism networks as well as molecular networks). We present stochastic simulations of idealized protein networks designed explicitly to carry out a numeric calculation. We explore the reliability of such computations and discuss error-correction strategies (algorithms) employed by living systems. Finally, we discuss some real examples of dynamic self-assembly processes that occur in living systems, and describe the RAM computer programs they implement. Thus, by viewing the processes of living systems from the RAM perspective, a far greater fraction of these processes can be understood as computing than has been previously recognized.