Friday, September 16, 2011

Epistemology of the Cell: A Review


The book, Epistemology of the Cell, by Dougherty and Bittner is a gem. Dougherty is a Professor at Texas A&M and has written extensively and brilliantly on the issue of using systems thinking in the growing field of genomics. Now systems thinking was, in way, started by Norbert Wiener in the 1930s, as he began to model various systems, from the dynamics of nerve fibers to the development of the first artificial arm. Yes, he did the arm, not the physician at Mass General. Wiener also developed the systems to the control the pointing of radar controlled guns on ships in WW II.

What made Wiener unique was that Wiener asked “Why”. The why meant finding the cause and the result and establishing the connection between the two, establishing the system. Physicists and chemists ask “why” questions, the find the cause and the causality chain. Biologists were for ages asking “what” as people who classified.

Darwin broke that chain of biologists somewhat by asking “why” as regards to evolutions and he is pilloried even to this day. Physicians often ask “what” and “how” and do not really want to be bothered by the “why”. It is not their jobs to find out why, just find out “what” is wrong and know “how” to fix it. Thus a physician is taught “what” to look for to diagnose prostate cancer and is also taught “how” to remove it. The physician does not know or care “why” the cancer is doing what it is doing. That is epistemological.

That, in essence, is a simplification of what Dougherty and Bitter go about to explain. Their book explores the challenges set forth to those exploring the gene and the cell and entices them to think beyond just the what and how to go to the why, the explanation of the process, the system, from cause to effect. The authors have written a treatise which compares a few others which look at the epistemological basis of research, asking the correct questions, and pursuing the best path to answer them (see the work by Winograd and Flores as a prime example written some 25 years ago). Dougherty has with his colleagues and associate been developing the ideas contained herein for well over the past decade and I have read many of his works, they provide great insight to what should be done.

Chapter 1 begins with a discussion of the definition of epistemology, on p. 2. They define scientific epistemology by what it addresses; knowledge and its truth. Then they use Kant to develop the transition from the Enlightenment to today. On p. 4 there is a brief but focusing discussion where he explains the simple Newtonian world maturing into an Einstein one and likewise a Watson and Crick paradigm of DNA/RNA/proteins into the way we understand cell dynamics today, as pathways, miRNAs, repressor enzymes and the complexity of both intracellular dynamics and intercellular dynamics. They rely extensively on a Popperian view of Science, which for those more familiar with Kuhn may tend to have some slight dissonance, but it holds together quite well.

All in all, Chapter 1 sets the stage well, both from establishing the necessity to have models which answer why, and to establish the counterpoint of the thinkers who have reverted back to Aristotelian classification as the finders of what.

Chapter 2 is a discussion of Aristotelian causality. p 14 states it well with the statement,

explanation must involve a causal relation…”.

On p 21 they state, “Galileo and Newton do not deny causality as a category of knowledge but they widen the scope of knowledge to include mathematical systems that relate phenomena, while bracketing “questions about nature” of the phenomenon…”.

On pp 33-34 they have an excellent discussion of Bertrand Russell’s work on causality.

In Chapter 4, on p 70, the authors use the work of Norbert Wiener and his associate Arturo Rosenbluth on Cybernetics. For it was indeed “the synergy of communications, control, and statistical mechanics…” that set the framework of how we should view cellular and organ dynamics. The authors then given examples of gene regulation. I would have simply stated that every cell and every organic system is a multidimensional distributed random process.

One could take a Feynman like approach and posit the obvious, and then fill in the details. The authors work from the bottom up to demonstrate their world view. Namely that when we look at cells we are looking at complex dynamic random systems. Systems we can ascribe states to, states being measurable quantities, which in turn operate on other states in a dynamic fashion.

In Chapter 5 the authors start the transition to complex state models. They again rely on the wisdom of Wiener on p 89 to state:

“Wiener recognized the difficulties that the mathematical requirement of science and translational science would present for medicine …”

For back in 1948 when Wiener had published Cybernetics, Medicine was still a “what and how” practice. It did not transition to a “why” approach. The translational science that the authors speak of is:

“… mathematical engineering, applied mathematics with a translational purpose..”

Namely, to translate nature to measurable quantities. Quantities which we can then by knowing the “system” we can then manipulate and predict. We can observe and we can control, the end goals of translational science. In a Popperian sense, the authors address the issue of measuring, predicting, and examining what does not do what we said it would.

The key arguments are developed in Chapter 9 and 10. Chapter 9 is the  “sola fides” discussion, faith alone, as a mantra to those who fail to understand the system nature of the cell dynamics. The authors, on p 149, evoke William Barrett, the insightful Columbia University philosopher, who wrote The Illusion of Technique, a superb work integrating the principles of epistemology and science in the late 1970s. Frankly, to see Barrett in a book of this type was an exciting surprise, for I had thought that Barrett was falling into obscurity, a loss to many who are struggling with issues that Barrett has thought through decades ago. The authors then on p 148 also discuss the nature of stochastic dynamic systems.

Dougherty brings insight via avenues that I found resonated strongly. The discussion on Wiener, where Dougherty, unlike Gleick in what I feel is presented with uninformed bias, sees Wiener as the father figure, one combining systems thinking, clear and built upon strong mathematical foundations, which is then integrated with real biological systems. Although I find their approach insightful and compelling, I would have taken pathways in cancer dynamics as somewhat well-defined stochastic systems.

For example, we know the effects of PTEN, the AKT pathway, and the MYC pathway, the p53 pathway, and the complex dynamics which are well described in the readily available NCI data base of pathways. One can use as states the concentrations of any one of these proteins and then state simply that they all interact with one another, the result being homeostasis or if a change cancer. The model is multidimensional, stochastic, highly complex, and strewn with “noise”, namely uncertainties. Models have been developed and tested for such cancers as prostate, melanoma and colon. Dougherty, himself, has made substantial contributions to this area. It would have been useful perhaps to demonstrate this approach as well.

On p. 163 the authors place a stake in the ground to say what would be expected for those to work in the field, that the books by Loeve and Cramer be used as standard bearers! As I read that in the book I looked on my bookcases and saw my old well-worn copy of Loeve, which got me through my PhD. Cramer was my core text for my introductory course, but then again it was MIT. Thus they set a high hurdle, but a necessary one for those to work in the field. My first book, Stochastic Systems and State Estimation (Wiley, 1972), in a sense was one of the many which established the bar.

The clear strike at the adversaries is set on p. 165. After again referring to Barrett and Kant, the authors end with:

“Does anyone really believe that data mining could produce the general theory of relativity?”

I think this can be extended. For example many researchers run millions of microarrays and are currently finding hundreds of SNPs or thousands of miRNAs and each time they send out a press release saying they have “discovered” some new “gene” or worse “cause” or “cure” of say prostate cancer or melanoma. In reality one does not know whether this is a marker for cancer, a marker for a predisposition for cancer or just plain noise.

What the authors, and others, have argued is that it is essential to have a well-defined dynamic system model of how say PTEN and AKT interact and how they in turn control MYC and where the controls on p53 are in this chain. The microarray analyses should be done in the context of defining the linkages in the state model and not as ends in themselves. The model can then be validated. From such a model we can then see conditions on their way to cancer and conditions representing advanced cancers. For example, recent authors have announced a way to measure PTEN in prostate cancer and laud that as a diagnostic step. In reality by the time PTEN has been deactivated there is most likely a metastasis. Understanding and refining a model is the essence of the “why” articulated by the authors.

On p 166 there is a superb critique of what the authors call the pre-Galilean thinkers, namely the biologists who like Linnaeus were really just classifiers of forms and shapes failing to understand why they were what they were. One must remember that biology was all too often just a study of things and a process of naming them and classifying them. The systems which made for these differences were little understood, and worse, beyond the mindset of many who practiced in the field.

Chapter 9 is. in my opinion. the pinnacle of their argument. Simply put, we should now begin to perform our experiments within the context of a model. For example, we know many of the pathways of the key genes intracellular, but we do not yet understand the dynamic model that controls them. Thus, when we do microarray tests, we should be doing them to determine the constants in the model and then validate that design. We should, in effect, identify the system, using a system framework, not just some unstructured set of classifiers. We have the structures, now is the time to put them to use.

Science is iterative. It is an iterative set of models and refinements. On p 171, the authors refer to Turing’s last paper on tessellation, or why zebras have stripes. The paper by Turing was submitted a day or two before he committed suicide and it was done without the benefit of Watson and Crick who were simultaneously doing their work at Cambridge. He intuited intercellular flow of some yet to be defined controlling substances. The concentration of these unknown substances would rise and fall in concentration and as such the color would change.

This approach has recently been applied, using a system model of flower genetics, and it explains and demonstrates the control of patterning in a genus of flowers. Having the model for this genus of flower, which is experimentally verifiable, one can then do the inverse, namely the controllability issue of creating desired flower patterns. That also is the essence in cancer dynamics, namely of creating a control or cure, but with a verifiable model. One must have the model, thus say the authors. Thus says nature! If one takes the authors systems approach and applies it to intercellular systems thinking, then it can be argued that the stem cell of cancer theory as has been recently evoked can be readily explained, as readily as those zebra stripes! That is the strength of the model posited by the authors.

My only negative is the price. There are some 187 pages and the price is well above $100. Amazon does have a better price but not much. That is over $0.60 per page, and for that I would blame not just Wiley, but the IEEE which somehow has all too inflated prices. This book is so insightful that the barrier to entry should be more modest. It is worth the insight at any price however.

And one last nit. The work of the authors should also shine a light on macroeconomics, which suffers not from a surfeit of models and mathematics, but from any ability to validate them, just look at the current state of the economy. Thus what biology brings from data to systems, macroeconomics brings from systems to reality. Both have to merge and be able to be predictive. We are approaching that in biology, especially with the insight of the authors, it is a pity that such has not yet passed the minds of those who opine on the world’s workings.