Can Networks Design Themselves?
A molecular biologist and a physicist at the Weizmann Institute of Science in Israel (see also 09/26/2003) wrote a paper in PNAS1 with an intriguing title: “Spontaneous evolution of modularity and network motifs.” Can a network arise spontaneously?
Biologists increasingly speak of the interaction of genes, proteins and metabolic processes in terms of networks (e.g., 12/20/2004, 03/22/2004, 01/28/2004, 01/27/2003, 01/10/2003). The networks with which most of us are familiar, like the power grid or internet, came about with intensive programming and intentional engineering. After the network architecture and the rules of interaction were defined, however, many unforeseen and spectacular patterns emerged. It could be argued that each emergent property of the internet had its roots in intelligent causes, however, since only sentient beings – humans – use the internet, and they do so with purpose and intent. In biological system there are also characteristic network-like patterns. Could these have arisen without purpose and intent? For Kashtan and Alon to prove this, they need to establish that networking behavior can be an emergent property of the molecules of the cell, without any programming.
In the history of computer software design, one important revolution was the invention of modular programming. Early programmers got tangled in their own “spaghetti code,” writing routines that jumped to other routines in such complex ways that the entire system became one single point of failure. Programmers realized that certain functions could be modularized, or segregated into independent routines that, though part of the big system, focused only on their own task. A module for addition, for instance, might take two undefined inputs, and have the function: “add these two inputs together.” The next module up the chain can call this module and give it any two numbers, and be assured the sum will be faithfully returned. Computer systems and networks built with a modular design were found to be much easier to maintain, and became much more robust against perturbations. A module could be upgraded or replaced without requiring a rewrite of the entire system. Biological networks also appear to work in modular fashion. Kashtan and Alon believe that they have found purely natural reasons for why this is so:
Biological networks have an inherent simplicity: they are modular with a design that can be separated into units that perform almost independently. Furthermore, they show reuse of recurring patterns termed network motifs. Little is known about the evolutionary origin of these properties. Current models of biological evolution typically produce networks that are highly nonmodular and lack understandable motifs. Here, we suggest a possible explanation for the origin of modularity and network motifs in biology. We use standard evolutionary algorithms to evolve networks. A key feature in this study is evolution under an environment (evolutionary goal) that changes in a modular fashion. That is, we repeatedly switch between several goals, each made of a different combination of subgoals. We find that such “modularly varying goals” lead to the spontaneous evolution of modular network structure and network motifs. The resulting networks rapidly evolve to satisfy each of the different goals. Such switching between related goals may represent biological evolution in a changing environment that requires different combinations of a set of basic biological functions. The present study may shed light on the evolutionary forces that promote structural simplicity in biological networks and offers ways to improve the evolutionary design of engineered systems. (Emphasis added in all quotes.)
In short, if the environment is modular, the network will become modular. This has been the problem, they reason, with computer models of evolution. Modelers used to give the computer a fixed goal and let the evolutionary algorithm figure out the way to reach it, by rewarding each routine’s “fitness” as it got warmer. These two researchers, instead, tried routinely switching the goal during the run. The networks that won out in the end were the modular ones:
The networks evolved under modularly varying goals were able to adapt to nearly perfect solutions for each new goal, within about three generations after the goal was switched. This evolvability was caused by the fact that the evolved networks for the different goals differed only slightly. For example, in many cases they differed in the threshold value of a single neuron, allowing switching between the networks with a single mutation….
Networks that evolve under modularly varying goals seem to discover the basic subproblems common to the different goals and to evolve a distinct structural module to implement each of these subproblems. Evolution under modularly varying goals produces networks that can rapidly adapt to each of the different goals by only a few rewiring changes.
So the winners evolved not only to be modular, but to be evolvable. This, they think, is the secret of how biological networks became so robust in spite of changing circumstances. Once a module for chemotaxis arose, for instance, a bacterium could reuse it with just a few “rewiring changes” if the chemical attractant changed. But who is doing the discovering? The subject of their sentence was, “Networks that evolve… discover…” The language of intent continues in another sentence in the ending discussion. Watch the subject:
In such cases [evolution with fixed goals], when the goal changes, the networks take a relatively long time to adapt to the new goal, as if it starts evolution from scratch. Under modularly varying goals, in contrast, adaptation to the new goal is greatly speeded up by the presence of the existing modules that were useful for the previous goal.
That last sentence used a passive voice verb: “adaptation… is greatly speeded up.” This hides the implication that the modules are seeking to adapt with goal-directed behavior. The authors are clearly not intentionally attributing intrinsic purpose to the modules. Their discussion of “fitness landscapes” in the subsequent paragraph treats the modules as pinballs on a bumpy landscape. Shifting goals keeps the landscape undulating so that the pinballs don’t get trapped on “local fitness maxima.” So is goal-effective modularity a true emergent property, as pointless and aimless as water running down a slope and seeking the least obstructed path? They actually experimented more to clarify this possibility. Notice the words information processing and useful:
One possible explanation for the origin of the motifs in the evolved networks is that modular networks are locally denser than nonmodular networks of the same size and connectivity. This local density tends to increase the number of subgraphs (42). To test this possibility, we evolved networks to reach the same modularity measure Q as the networks evolved under modularly varying goals, but with no information-processing goal (see Supporting Text). We find that these modular networks have no significant network motifs (Fig. 9). They show relatively abundant feedback loops that are antimotifs in the networks evolved under modularly varying goals. It therefore seems that the specific network motifs found in the evolved networks are not merely caused by local density, but may be useful building blocks for information processing.
In other words, unless information processing was programmed in as a goal, mere environment-shifting produced anti-motifs – a backward step. That is why their only success came with emphasis on achieving useful building blocks for “information processing.” But what is “useful” to a network? Why would a non-sentient network seek to process information? If not the network, is there an outside agent that cares about such things? Like the tree in the woods falling without a sound, can there be “information” without a mind to conceive of it?
At this point, they compared their computer models to actual biological networks. Here, they could not escape portraying the genes and cells as if they were tiny sentient beings:
How is evolution under modularly varying goals related to actual biological evolution? One may suggest that organisms evolve in environments that require a certain set of basic biological functions….
[They discuss chemotaxis evolving as the chemical attractant changes.] When environments changed, these modules adapted over evolution to sense and chemotax toward other nutrients. Had evolution been in a fixed environment, perhaps a more optimal solution would have mixed the genes for these different tasks (e.g., a motor that can also sense and transport the nutrient into the cell), resulting in a nonmodular design….
An additional biological example occurs in development. Different cells in the developing embryo take on different fates. Each cell type needs to solve a similar set of problems: expressing a set of genes in response to a given time-dependent profile of a set of extracellular signals. However, in each cell type, the identity of the input signals and the output genes is different. Thus, in development, cells need to perform essentially the same computations on varying inputs and output: a modularly varying goal. The solution found by evolution is a modular design where signal transduction pathways (such as mitogen-activated protein kinase cascades), which are common to many cell types, hook up to specific receptors and transcription factors that are cell type specific. This design allows simple rewiring of the same pathways to work with diverse inputs and outputs in different cell types. Over evolutionary time scales, this design allows the addition of new cell types without the need to evolve dedicated new pathways for each input and output….
They threw in a bonus that their study might help engineers “evolve” improved networks. But understanding biology was clearly the intent of the paper. How to get biological design without a designer – that quest was evident in their last two sentences. “In summary,” they said, “this study presents a possible mechanism for spontaneous evolution of modularity and network motifs. It will be important to extend this study to understand how evolution could generate additional design features of biological systems.”
1Nadav Kashtan and Uri Alon, “Spontaneous evolution of modularity and network motifs,” Proceedings of the National Academy of Sciences USA, published online before print September 20, 2005, 10.1073/pnas.0503610102.
Foul, time out, game over. They just violated the No Free Lunch Principle. You just caught them in the act. This is the persistent sin of evolutionists, engraved with an iron stylus on their stony hearts. They only get away with this evil because no preacher is allowed past the walls of the Darwin Party fortress to call them to repentance. Naturalists cannot attribute will, purpose, intent and information processing to mindless entities. This violates their core assumptions as philosophical naturalists (materialists), whose goal was to rid natural explanations of teleology (purpose, final causes).
William Dembski in his writings, especially the book No Free Lunch, drives home the point that “no evolutionary algorithm is superior to blind search” – that is, unless information has been smuggled in behind the scenes. Consider his famous treasure island analogy. If you are on an island where treasure is buried, and have no clues, blind search is your only option – a very inefficient method, becoming more hopeless as the size of the island increases. A friend tells you there is a treasure map. Encouraged by this new hope, you go to the hostel where the map is locked in a cabinet. You find, to your despair, that there are a million treasure maps, all different, all claiming to be the right one. In a real sense, you now have moved your blind search to another space: the space of maps, where the the correct map is the new treasure. This analogy can be extended indefinitely: your friend says a guru knows which map is the correct one. You go to the mountain top, only to find again, to your despair, that a million gurus await you promising you the path to enlightenment. The only way out of this infinite regress is to get true, useful information from someone with knowledge of the treasure’s location. Anything else is blind search.
Now, to a Darwinist, who is going to provide that information? Surely not the environment. Surely not random strings of DNA. Surely not randomly floating bits of protein. None of them can possibly have any goal or purpose in mind, or any embedded knowledge of the best way to build modular networks that grow, reproduce, and function robustly in changing environments, complete with error-checking, coded instructions. We must emphasize this point: any attempt by a Darwinist to impose wish fulfillment, goal-directed behavior, or teleology on these molecules is strictly forbidden. One must visualize these molecules as completely and utterly indifferent to success or failure. They care nothing if a function is achieved, and nobody is there to cheer them on. Kashtan and Alon conveniently left the origin of any primitive network as an unsolved problem. Fine; they still must maintain the impersonality of that initial network. By analogy, picture a bunch of unthinking robots that had an initial purpose imposed on them by some unexplained inventor – say, to sort and stack rocks. Get real, now, and ask yourself: realistically, is changing the environment going to improve their modularity and evolvability? If you come back after the magic factor of “evolutionary time scales,” will you expect to see the robots building airplanes, printing books and conducting orchestras? Of course not. Remember, the robots are not sentient beings. They couldn’t care less whether some new “function” emerges, or whether they rust in a colossal heap of rubbish (a more thermodynamically favored outcome).
Where evolutionists cheat incorrigibly is by personifying molecules into purposeful entities, or by invoking mindless processes as creative agents. Notice again how subtly they do this: “One may suggest that organisms evolve in environments that require a certain set of basic biological functions….” This “suggestion” makes no sense unless one personifies the environment as a manager setting design requirements, and the evolving entity “needing” or “wishing” to fulfill them. Whether invoking Tinker Bell with her mutation wand, or shuffling environments to get the desired outcome, evolutionists are playing the guru telling the treasure hunter which map is the correct one. This is forbidden. The only evolutionary algorithm that is permissible on Darwin Island is blind search. The island is the size of the universe, and according to our online book, the chance of getting even one useful protein, let alone a modular, functional network that can adapt to changing environments, is less than Dembski’s “universal probability bound” of 1 chance in 10150. It would be easier for a blindfolded man to pick a single marked penny out of a whole universe packed with pennies than to expect chance to succeed at this task. But chance is all they have according to the No Free Lunch Principle. And no, they cannot cheat by saying “Well, we are here, therefore it must have happened somehow.” Unless they are willing to consider intelligent design, this is a post hoc fallacy.
In conclusion, Darwinian materialism must retreat into pantheism, or else give up in despair. You just read two Darwinists whispering about biological entities as if they were sentient beings. Their pantheism is implicit, despite their intent to explain biological networks in materialistic terms. As such, they are teaching pantheistic nature religion, not science – and of all places, right there in central Israel! This is right where Baal worship, another nature religion, was condemned by the Hebrew prophets 2800 years ago. Where is Elijah when we need him? Phillip Johnson? Henry Morris?