Putting intentions into cell biochemistry: an artificial intelligence perspective.

The living cell exists by virtue of thousands of nonlinearly interacting processes. This complexity greatly impedes its understanding. The standard approach to the calculation of the behaviour of the living cell, or part thereof, integrates all the rate equations of the individual processes. If successful extremely intensive calculations often lead the calculation of coherent, apparently simple, cellular "decisions" taken in response to a signal: the complexity of the behavior of the cell is often smaller than it might have been. The "decisions" correspond to the activation of entire functional units of molecular processes, rather than individual ones. The limited complexity of signal and response suggests that there might be a simpler way to model at least some important aspects of cell function. In the field of Artificial Intelligence, such simpler modelling methods for complex systems have been developed. In this paper, it is shown how the Artificial Intelligence description method for deliberative agents functioning on the basis of beliefs, desires and intentions as known in Artificial Intelligence, can be used successfully to describe essential aspects of cellular regulation. This is demonstrated for catabolite repression and substrate induction phenomena in the bacterium Escherichia coli. The method becomes highly efficient when the computation is automated in a Prolog implementation. By defining in a qualitative way the food supply of the bacterium, the make-up of its catabolic pathways is readily calculated for cases that are sufficiently complex to make the traditional human reasoning tedious and error prone.