“Reductionism: Instagram approach to science”

One of the greatest strengths, and one of the greatest failings, of modern science is reductionism. Reductionism has allowed us to dissect and understand some of the most important natural phenomena. Some would argue that reductionism is at the heart of the Scientific Method. Marvin Minsky has said, “In science one can learn the most by studying the least.” This is true in many instances. However, one can understand the least by studying the least as well.

This approach is exemplified by the molecular approach to biology. I must admit, for example, that as a biology undergrad in the 1980s, I studied almost nothing larger than a cell. Neither did virtually any of my friends. When genetic engineering was invented in the 1980’s, the rest of biology came to a screeching halt. Almost every biologist dropped whatever they were studying and switched to molecular biology. For example, my advisor in college was a professor who had started as an insect biologist, but then switched almost completely to molecular biology.

This was, in some ways, a major advance for biology. It capped a long road that biologists traversed to become “real” scientists, from being considered “natural historians” or “naturalists,” scientifically inferior to “natural philosophers” like chemists and physicists. Only a few would dispute that biologists are now as much of scientists as physicists. In fact, we have mostly forgotten the fundamental difference between natural philosophy and natural history, even though at its heart, biology is a very different kind of science that the hard sciences like chemistry and physics. Biologists are still in some important ways historians as much as we are scientists, because provenance matters in biology.

Reductionist approach is fantastic, in that it allows us to isolate one part of a biological phenomenon and really understand it. We can take a single gene, pull it out of a cell, and study it in complete isolation. This permits us to study the gene without the confounding factors. It allows us to “freeze frame” natural phenomena and study it carefully.

However, the molecular biology tools that allowed biologists to graduate into the top-tier of science carried with it a dear price. The central tenet of reductionism, that you can understand something solely by understanding its parts, is fundamentally flawed.

The problem is that a part of something is not the same as the entire thing. When you study only a small part of a system, you often miss what’s known as emergent phenomenon. The classic example is water vs. a water molecule. A single molecule of water is not solid, liquid, or gas. It’s only when you have many molecules of water that you have such a thing as a solid, liquid, or gas phase. Similarly, when you study a single kinase, it’s difficult to understanding emergent phenomena like hysteresis, system resilience, and system redundancy–for example, will inhibiting the kinase result in rapid upregulation of alternate pathways? You really need to study the entire cascade. Don’t get me wrong–it’s important to understand the enzyme and how it functions, but it is also important to understand the system as well.

To take another example, a body is not the same as merely a collection of cells. You can’t understand the circulatory system by only studying cells. You can’t understand the nervous system by study only neurons. You need to understand how the neurons are connects to each other, for example. There are properties of the circulatory or nervous system – such as endothelial dysfunction or insulin resistance what can’t be described on a cellular level. Indeed, they can’t be described without reference to other bodily systems. In short, emergent phenomena are critical to biology.

Lack of awareness of emergent phenomena is what often leads to dead ends in drug development. For example, we often try to study enzymes or signal molecules in isolation. We are sometimes surprised when we try to modulate those targets and find that the results are opposite of what we expect. We block a molecule, for example, and see an increase in its biological effect, for example. This is often because the other parts of the system respond to the intervention.

This brings us to a second point: you can’t fully understand biology if you ignore the spatial properties. I will coin a new term, spatio-emergent phenomena, for this. You can’t isolate a kinase from a cell, or a tumor cell from its extracellular matrix, and hope to understand it completely. Tumor cells in a tumor cell line is not the same as a tumor cell in a tumor in a body (for many reasons). Studying a gene in isolation, can be powerful, but also misleading, in that the genetic milieu can influence the function of the gene.

It turns out, for example, that the specificity of many kinases and probably other enzymes, is dependent not only on its structure but on what subcellular compartment of the cell it localized to. There are specific anchoring proteins that localizes the kinases to the subcellular compartment. Without this spatial information, our understanding of kinase biology is incomplete. Similarly, when immune cells mature and get programmed, they pass through different areas of the lymph nodes, and there is information conveyed to the immune cells in the different spatial areas of the lymph nodes.

In continuing in this vein, let me then bring up a third concept: temporo-emergent phenomena. We also can’t fully understand biology if we ignore the temporal aspects of systems. As an example, many hormones undergo cyclic variation. You can, for example, treat prostate tumors by administering either an androgen agonist or an antagonist. This is because the cyclic up and down beat of androgens is what stimulates prostate growth. If you keep the androgen levels high or low, without variation, you will inhibit the growth.

So, I think I’ve made a reasonable argument why we can’t ignore emergent phenomena if we want to fully understand biology, or to do drug development. Let me take one example to illustrate the point: biological cascades.

One common approach to dissecting a biological system is to perform an expression analysis. We might look at the expression profile of genes or proteins, and try to find genes or proteins that are expressed differently between normal and diseased people or animals.

This is a classical reductionist approach. Treat all the genes/proteins equally, and home in on the genes that are the most upregulated. The genes/proteins that are the most changed become the targets or interest. Very democratic.

The problem is several-fold.

First, many biological systems are cascades, like the coagulation cascade or the inflammatory cascade. The genes/proteins at the bottom of the cascade are the effector proteins, and they often have the most noticeable change in abundance. The problem is that the most important genes/proteins at the top of the cascade are often the most important, but often have the lowest expression levels. For example, epo is present in very small amounts in the blood, but is one of the most important proteins in the body. It is present in such small amounts because it is so potent and so important.

Second, the expression analysis often don’t examine the timecourse of the expression. To isolate the most important molecules, it is advisable to characterize the timecourse because the more important molecules often tend to be activated or expressed first.

Third, the expression analysis often don’t examine the robustness of the response. There are often multiple feedback loops and alternate pathways that insure that the cascade proceeds, and the different loops sometimes have different robustness.

Fourth, the expression analysis typically don’t examine the countercascade system. It is often assumed that the biological cascades self-extinguish once the initiating stimulus disappears. however, this is usually not the case. In most systems where the countercascade system has been examined, there is an active feedback system that turns off the cascade. The cast-off peptide from prothrombin and fibrinogen appears to have counter coagulatory effects, and some of the end products of the arachidonic acid pathway seem to have anti-inflammatory effects. Electrical engineers tell me that the most important part of controlling a circuit is controlling the feedback mechanism, not the forward activation part of the system. It is possible that the most important way of controlling many biological systems is by controlling the feedback mechanism. Unfortunately, we as biologists are trained to look at effector functions, not control functions

There are a few companies taking a system approach. Merrimack, for example, found that ERB3, which based on its abundance on the cell surface would not typically be considered to be important driver of tumor growth, found that once the system was understood and modeled. it was a key driver for growth.