The imitation of living and sentient beings by machines is recently much on the minds of many Americans.  A computer designed and built by scientists and engineers at IBM, “Watson,” convincingly defeated two former “Jeopardy” champions in a televised competition on the long-running game show.  This triumph of a machine over humans has stimulated both recollections of the last hallmark event in this series, the defeat (in six games in May of 1997) of chess grand master Garry Kasparov by another IBM computer, “Deep Blue,” and speculation about the future of artificial intelligence.  

Interest in the ability of computers to simulate human thought and intelligence persists in parallel with the tendency of many biologists and biomedical scientists to think of biochemical entities, (such as transcription, splicing, and signaling complexes), cells, and whole organisms as analogous to machines.  For example, the hijacking of cellular processes by viruses often invokes a phrase referring to the exploitation, by the virus, of cellular “machinery.”  Biologists frequently refer to cellular structures, like ribosomes, as “molecular machines.” 

The late Robert Rosen, a professor of physiology and biophysics at Dalhousie University, addressed the machine metaphor for organisms in his challenging and unusual book (1991) exploring the key attributes of living systems that distinguish them from non-living systems.  Rosen offered a thoughtful critique of conventional biological thinking that was motivated in part by his knowledge of mathematical logic, possibly rendering his perspective less than maximally accessible to the typical biologist or biomedical scientist.  

The core of Rosens’ thesis is that the fundamental assumptions behind the machine metaphor are profoundly flawed.  One of the key assumptions is that, like a watch or a television, a cell or organism consists of a set of clearly-delineated parts into which the cell or organism can be decomposed and whose individual properties will be sufficient for explaining the physiology (broadly construed) of the cell or organism.  Rosen maintained that there is no such easily-identified set of parts such that their individual properties will account for the behavior of the cell or organism.  He noted that physics, at least at the time of his writing, provided an example of a relatively simple system (the “three-body problem”) for which analysis based on the properties of the readily-identified individual components failed to provide a reliable, predictive account of the behavior of the whole system.  

I have just published (2011; available online ahead of print) a commentary that examines complexities associated with mapping functions onto molecular structures, whether genes or gene products.  One of my central claims is that the functions attributed to genes and gene products are often influenced strongly by both internal and external contextual factors.  So, for example, the effects of a cytokine on the pathogenesis of an inflammation-related disease can depend on when in the process the expression of the cytokine is being varied (Jacob et al., 1989). Similarly, the effects of the expression of particular genes on the process of carcinogenesis can reflect the stage of tumor development (Singh et al., 2010) or other factors (Miao et al., 2009).  

Douglas Hofstadter (1979) has pointed out that, “In geometry, the words “point”, “line”, and so on are undefined terms, and their meanings are determined by the axiomatic system within which they are used.”  Somewhat analogously, the functional significance of any gene in a genome may depend, and sometimes to a readily discernible extent, on the alleles present at other loci distributed throughout that genome.  Thus, attributions of in vivo functions to molecular-scale structures in particular circumstances while reasonable and often valuable cannot be regarded, in the general case, as definitive. 

That the correlation of genotypes with phenotypes should be so challenging is, in my view, a predictable consequence of the evolutionary process.  While it is also possible to speak of the evolution of technology and machines there are critical differences in the evolutionary processes for animate versus inanimate entities.  For example, the composition of genome-encoded biological components in terms of numerous independently variable subunits permits an endless tweaking and fine-tuning, “tinkering” in the apt phrasing of Francois Jacob (1982), that proceeds spontaneously to an extent that is not duplicated by machines with perhaps rare exceptions.  Furthermore, organisms are, so far at least, necessarily directly derived from other organisms.  Such direct and comprehensive material continuity does not generally apply to machine “reproduction.”  

Thus, the intricate and complex relationships found among the parts in a machine are so far at least, less intricate and less complex in a few key respects than the relationships found among the parts in a cell or organism.  Perhaps, machines will some day be more like organisms, but then they will have transformed into entities that are not precisely what we mean by “machine” in current usage.  In the meantime, the machine metaphor for living systems should be used sparingly, with caution, and in precisely-defined ways to minimize the risks of flawed patterns of thought. 

References 

Rosen, R. Life Itself: A Comprehensive Inquiry into the Nature, Origin, and Fabrication of Life. Columbia University Press, New York, 1991. 

Greenspan NS. Attributing functions to genes and gene products. Trends Biochem Sci. 2011 Jan 24. [Epub ahead of print] PubMed PMID: 21269834. 

Jacob CO, Holoshitz J, Van der Meide P, Strober S, McDevitt HO. Heterogeneous effects of IFN-gamma in adjuvant arthritis. J Immunol. 1989 Mar 1;142(5):1500-5. PubMed PMID: 2493048. 

Singh MK, Izumchenko E, Klein-Szanto AJ, Egleston BL, Wolfson M, Golemis EA. Enhanced genetic instability and dasatinib sensitivity in mammary tumor cells lacking NEDD9. Cancer Res. 2010 Nov 1;70(21):8907-16. Epub 2010 Oct 12. PubMed PMID: 20940402; PubMed Central PMCID: PMC2970659. 

Miao H, Li DQ, Mukherjee A, Guo H, Petty A, Cutter J, Basilion JP, Sedor J, Wu J, Danielpour D, Sloan AE, Cohen ML, Wang B. EphA2 mediates ligand-dependentinhibition and ligand-independent promotion of cell migration and invasion via a reciprocal regulatory loop with Akt. Cancer Cell. 2009 Jul 7;16(1):9-20. PubMed PMID: 19573808; PubMed Central PMCID: PMC2860958. 

Hofstadter, D. R. Gödel, Escher, Bach: An Eternal Golden Braid. Basic Books, Inc., Vintage Books Edition, New York, 1979, p. 456.

Jacob, F. The Possible and the Actual. University of Washington Press, Seattle, 1982.