Tuesday, March 13, 2012

Putting the Reductionist Back Together Again



For the past century, at least, a debate has raged in the poster pinned halls of scientific institutions over two methods of investigation; namely, top down or bottom up.  These two methods can be seen in the text books published in many scientific fields.  In my own field of neuroscience the text books are divided into molecular neuroscience and systems neuroscience.  The bottom-up approach, also known as the reductionist model of investigation, focuses on sub-cellular and cellular mechanisms in order to describe biological phenomena from cilia dynamics in the gut all the way up to group behavior patterns.  However it is limited in many ways by the techniques used in these investigations and is criticized for an inability to predict complex systems accurately. 
While it has been the job of reductionists for hundreds (perhaps thousands) of years to take things apart in order to investigate them, it is now time for them to prove that they can put everything back together.  Some think that this can’t be done (see the viewpoint by Van Regenmortel in EMBO reports).  Indeed, it is believed that systems create what are known as emergent properties.  While this may be true, I believe that given enough information we can predict these emergent properties via a bottom up approach.  These emergent properties, are not, as some believe, unpredictable solely by looking at the parts, instead they are proof of our ignorance about the system, primarily network dynamics, and they do not just occur in biological systems.  The internet is one example of a man-made system that totally surpassed the expectations of its creators.  Two of its main functions; email and the World Wide Web, were not predicted by its inventors (See discussion in the book Linked, by Barabasi, pg. 149).  Although some would argue that the internet itself is an emergent property of a biological system as well.
The universe is based on rules, those that we know and those that we have yet to discover.  The failure of the reductionist method is primarily a failure to predict the complex interactions between its simpler parts.  Above, I mention the work of Barabasi, who has studied the formation of networks, (see also my Training for our Future post) and found that networks, whether they are found on the web, between actors in Hollywood, or within the molecular pathways of cells seem to follow a stereotyped pattern.  Could universal principles such as network algorithms be applied to reductionist finding to help predict emergent properties in these systems as well?  
Graded and time separated exposure to various external stimulants at the cellular level leads to a sequence of events important for lineage specification in the developing organism.  An example of this has been well studied by the motor neuron development specialist, Tom Jessell, who has worked to characterize lineage progression in motor neurons through their gene (primarily the Hox gene family) and protein expression patterns; events that are dictated not by the cells alone but by their environment.  Arthur Kania, who worked as a postdoc in Jessell’s lab, published a paper in Neuron last year that suggests not only that the levels of receptors are important in the development and diversity of axon guidance but the sub-cellular distributions and relative ratios between their binding partners are important as well.  It has been suggested that the complementary  co-expression of ligands and their receptors on the same cell may also modulate the ability of cells to respond to molecular signals.  It is likely that this type of complex molecular interplay is built into cellular interactions at every stage of life.  Some preliminary examinations of the literature suggest the Eph receptors and ephrins, the cell surface molecules examined in the Kania study, have similar variations in expression levels in many cell types.  The Kania study suggests a way in which a simple repertoire of ligands and receptors can lead to a diverse cellular responses.  Furthermore, other receptors have variations in expression at different stages within their lineage development or based on their location in the organism which underlies two main ideas. One; that the cellular environment affects the function of surrounding cells, and two; that cells have the plasticity to differentiate due to the up or down regulation of surface receptors.  I suspect that an analogy to this interplay between individual cells and their micro-environment can be found in other complex systems as well.  The examination of the dynamic interplay between cells will help to elucidate the missing steps between the parts and the emergent properties.
Graded responses to complex stimuli and variable protein expression patterns in individual cells within a cellular system underlie the frustration reductionists have in explaining emergent properties.   However, the work of Kania, Jessell and others suggests that complex mechanistic understanding of a system can be obtained via an in depth examination of the parts; which is the definition of the reductionist approach.  Furthermore, examples such as genome wide association studies and high throughput molecular screening, suggest that information acquisition and processing technologies may allow scientists to examine variables too complex to decipher in previous generations.  The coupling of our age old obsession with taking things apart and our developing understanding of network dynamics, might allow the reductionists to finally put the puzzle back together. 

No comments:

Post a Comment

As always, this blog is also a forum, discussion is appreciated.