Complexity in Biology: an Epistemological Nightmare

Given the massive amount of research in biomedicine it is tempting to think that we are now in the process of completing the picture of biological processes and just need to keep on moving forward. I wish, a more sober conclusion is that ongoing research into the particulars of biological processes and is going to leave us further bewildered. Fortunately there are some researchers who are strenuously attempting to create models of biological processes This particular approach offers considerable hope because it suggests that we do not need to understand in detail all the relevant processes in a biological function. You can read the Science Daily News item here. The authors even go so far as to state:

 In particular, these findings suggest not only that one may not be able to understand individual elementary reactions from macroscopic observations, but also that such an understanding may be unnecessary.

For myself at least this is very promising because I have long held the view that while the research into the individual elements involved in biological processes is very important a more useful understanding will never arrive simply through more "butterfly collecting" but through the development of models that do not require a taking into account of all these individual agents. To understand why I have long held this view please read on.

 The "bottom up" approach  is fundamentally limited

Too many elements


A long time ago the gifted French mathematician Henri Poincare came up with the "3 body problem". In essence the argument is that it is impossible to calculate gravitational impact on the trajectories of 3 massive bodies. This is indicative of a general computational problem in that when confronted with multiple agents it becomes extremely difficult and often impossible to know the final outcomes of these agents. In biological processes this problem becomes more acute because many biological agents have context dependent roles.

Even if we could measure all the relevant agents in a given process chances are we would be left with a plethora of data but little advance in our understanding of how all these processes give rise to specific outcomes. We have mountains of data, I suggest if the answer is to be found in creating more data we may well end up missing the forest because we are too focused on the trees.

The measurement problem

It is painstaking work to determine the influence of one biological agent. Typically many experiments are required and it is not that unusual to come up with paradoxical results. We have enough trouble just determining the impact of one biological agent yet what is required is the ability to simultaneously measure the impact of a number of biological agents in a given process. Why? Because as I previously stated the impact of any biological agent is context dependent. Measuring the effect of one agent in isolation is not a representation of what actually happens in biological processes. Unfortunately it is not experimentally possible to measure all the relevant effects of the biological agents in a given process. There have been some small steps in that direction but again the answer is not going to be found through such a research strategy.

Context Dependent Function

I'll illustrate this by way of example. We've all heard about those nasty inflammatory mediators and how much damage inflammation can cause. There is a great deal of research into inflammation, so much that one is left with the impression that inflammation is evolution's nasty trick to hasten our demise. The key inflammatory mediator is interleukin 1. Time and again studies will show how elevated levels of this molecule increase the risk of dementias, cancers, heart disease, autoimmune disorders, and various inflammatory related conditions.

Within the mountain of research into interleukin 1 though there are many research items that point to interleukin 1 having critical roles in repair processes, growth processes, and there is even research indicating that interleukin 1 may be important in maintaining Long Term Potentiation, often regarded as essential to long term memory storage.

Within any biological process the impact of a given agent is often contingent on the presence of other agents and the general physiological state of the organism. Thus while, for example, a carbon atom is a carbon atom is a carbon atom, in biological processes interleukin 1 can be a dangerous inflammatory mediator but can also be important for tissue repair etc. I will even go so far as to suggest that interleukin 1 may be very important in enabling the removal of amyloid, the key protein involved in Alzheimers, thereby reducing the risk of dementia.

Can We Still Dream?

Yes, but only if we stop clinging so desperately to the 19th century reductionist dream that all we have to do is gather sufficient data and the solution will reveal itself. The history of science offers some support for this notion but there is also plenty of evidence that it was hard thinking on the basis of available evidence that led to the creation of models providing greater insight into what is going on.

The "butterfly collecting" must continue, we still have a great deal to learn about the qualities of specific biological agents and their interaction with other agents. BUT it will never be sufficient in explaining biological processes. At some point we must be prepared to boldly enter into a strategy that requires fresh thinking about modeling biological processes. There is promising research into this area. I hope that the previously mentioned paper represents a cardinal breakthrough in addressing this challenge because it overcomes all the problems I outlined above. To be honest though, I"m not holding my breath.