In two previous posts (here and here) I noted that MIT bioengineer Dr. Erika DeBenedictis had argued in a TEDx talk, “It’s Time for Intelligent Design,” that biological systems are the result of “random chance” and as a consequence “imperfect.” She goes on to argue that the human genome contains a “dumb mistake” that is “painfully embarrassing” because she thinks “any human engineer” would catch this and “no one would engineer this deliberately.” 

But is there another way to look at things? Yes! Biology is actually a showcase of elegant designs with displays of optimality in core infrastructure. This optimality raises the bar for random mutation and natural selection and raises the possibility that not only may biology reflect good design but in many cases it may be the best design given physical constraints. Discovery of optimality in biology urges caution in labelling “poor design” in unclear cases because optimality suggests good overall design. Add to this the history of embarrassing “poor design” enigmas in biology stripped of their pomp by recent research and all that remains is a reminder that overall biology appears very well designed and unaccounted for by evolutionary explanations.

Perhaps the biggest mistake in first impressions of biological design is to overlook constraints. Although many biologists have little exposure to the idea of trade-offs, biology has many constraints. Self-replication, physical laws, chemical properties, and interacting ecosystems all have their own demands. Each of these broad categories has beneath its surface a plethora of subcategories. Understanding constraints is key to determining if a design is good for the intended function. Without considering all relevant constraints and empirical testing it is not possible to say whether an aspect in biology is poorly designed. But when scientists have made assumptions about function, examined some design constraints, and done empirical testing or simulations, they discover optimality in core biological infrastructure. This wrecks the idea of overall poor design in biology and leaves us grappling with the question: Is natural selection the best explanation for the type of design we see in biology? Let’s look now at  some specifics on optimality in core biological infrastructure. I will begin with an example from embryology, then turn to metabolism, and finish with a discussion of the breadth of chemical space covered by the natural amino acids. 

An Example of Optimality from Embryology

Embryology is the study of how organisms develop. Optimality observed in development suggests good design at a foundational level. At Princeton, William Bialek has been studying how organisms approach physical limits of precision in their development. He likes to create a theoretical idealization and then see how biological systems measure up:

Here, we test the hypothesis that the fly embryo achieves an optimal decoding of position given access to the gap gene expression levels in each individual nucleus, at a single moment in time. While optimality is a controversial hypothesis, we emphasize that, in the present context, it makes unambiguous, quantitative predictions, which we test. 

(Petkova et al.)

His group’s research has demonstrated that the GAP gene expression data in Drosophila is approximately optimally decoded from a mathematical perspective to provide each cell a unique identity in a developing embryo.

Perhaps the most important qualitative conclusion from our results is that precision matters. We are struck by the ability of embryos to generate a body plan that is reproducible on the scale of single cells, corresponding to positional variations ∼1% of the length of the egg. As with other examples of extreme precision in biological function, from molecule counting in bacterial chemotaxis to photon counting in human vision, we suspect that this developmental precision is a fundamental observation, and to the extent that precision approaches basic physical limits it can even provide the starting point for a theory of how the system works. 

(Petkova et al.)

Observing such precision in development makes DeBenedictis’s proposition, “Biology is imperfect. Let’s make it better,” seem like an overstatement. Given that “the embryo can extract all the available information about their position, but only if the concentration measurements approach the physical limits of information capacity” (Bauer et al.), the most straightforward general statement is that the development of fruit flies is optimal or the best possible design. This research raises the additional question as to the extent that other organisms may reach the physical limit of information capacity in their development. While more research is required to determine the extent of this optimality in development across the biological kingdom, it is not illogical to expect similar precision in other organisms that have equal capability and need for precision in development.

An Example of Optimality from Metabolism

If one is committed to biology being the product of “4 billion years of random chance,” you will anticipate poor design. German biophysicist Reinhart Heinrich did. Notice first his commitment (sentence 1), then his prediction (sentence 2), and finally his surprise (sentences 3 and onward) that mutations in metabolic systems result in worse biological function. Sentence numbers have been added to the quote below:

[1] [T]he biochemical reaction networks are the result of natural selection where the contemporary metabolic systems have been developed by a stepwise improvement of the functioning of the different subsystems of the cell. [2] Certainly, this development did not lead to a ‘global optimal state’. [3] However, it is an experimental fact that a change of the kinetic parameters of enzymes by mutations in contemporary metabolic systems mostly results in a worse biological function. From this, we may conclude that with respect to present-day metabolic systems we are confronted at least with a “local optimum”.

(R. Heinrich, emphasis added)

It is time to make better predictions! Glycolysis, which is the central energy pathway for all living systems, has been shown to provide maximal ATP at minimal protein production expense after a consideration of several constraints (Ng et al.). In E. coli glycolysis appears to be the shortest pathway to produce all essential biomass precursors (Noor et al.). Another analysis showed that this pathway avoids toxic intermediates and prevents leakage by phosphorylation of the intermediates (Bar-Even et al.) The second half of glycolysis was shown to be able to withstand the greatest amount of flux compared to alternatives (Court et al.). Measurement of metabolite concentrations and flux has also uncovered that free energy is well partitioned to prevent backward flux (Park et al.). 

Despite these exciting optimality discoveries for glycolysis, a 2014 biology textbook remains focused on describing glycolysis as not all that great. Note how these authors do not update students on the strengths of the pathway. Instead they emphasize that evolution is an incremental process:

Why does glycolysis take place in modern organisms, since its energy yield in the absence of oxygen is comparatively little? There are several possible answers. First, the process is energetically efficient and better than the alternative — no ATP. Second evolution is an incremental process: Change occurs by improving on past successes. In catabolic metabolism, glycolysis satisfied the one essential evolutionary criterion — it was an improvement. 

(Mason et al.)

Research has now demonstrated that there is a trade-off for making more ATP. Consideration of this trade-off and thermodynamic feasibility over a range of intermediate and ATP/ADP concentrations has updated glycolysis “status to be ‘Pareto optimal.’” Pareto optimal is the state wherein no single criterion can be made better without making something else worse. Here’s what they say:

Pareto optimality analysis between energy efficiency and protein cost reveals that the naturally evolved ED and EMP pathways are indeed among the most protein cost-efficient pathways in their respective ATP yield categories and remain thermodynamically feasible across a wide range of ATP/ADP ratios and pathway intermediate metabolite concentration ranges. 

(Ng et al.)

To make my point, I reiterate that glycolysis, as the best understood metabolic pathway, was initially thought to be suboptimal probably due to a “4 billion years of random chance” commitment, but has since been shown to be optimal. This should encourage future research and similar constraint considerations for other metabolic pathways to determine the extent of this optimality in biology. I also encourage acknowledgement of pareto optimality for glycolysis in all future biology textbooks. 

An Example of Optimality in the Natural Amino Acids

The natural set of amino acids has been studied for its unusual properties and determined to be exceptional. Perhaps unaware of this, DeBenedictis chose the natural amino acids as an example of imperfection in biology. She says in her TEDx talk:

One of the big limitations of biology are the basic building blocks themselves. There are only 20 different types of amino acids. And when you look at the amino acids, chemically speaking, they are actually kind of boring. 

This statement doesn’t reflect an appreciation that the natural set of amino acids is considered by many scientists to be “a largely global optimum” for the needs of aqueous biochemistry. The natural set of amino acids has been said to possibly “represent a largely global optimum, such that any aqueous biochemistry would use a very similar set.” (Ilardo et al.)

Sets that cover chemistry space better than the genetically encoded alphabet are extremely rare and energetically costly. 

(Ilardo et al.)

Three properties of amino acids, namely size, charge, and hydrophobicity, are the primary constraints studied so far for the natural set of amino acids (Mayer-Bacon et al.; Mayer-Bacon and Freeland). Given only those three properties, one study found that out of 10 million sets of size 20, only 6 sets out-performed the current set. (Mayer-Bacon et al.) A glance at one of those 6 “better” sets reveals that consideration of hydrophobicity, size, and charge was inadequate to produce a set compatible with current protein structure:

This xenoalphabet exhibits superior coverage under the current definition of the term, but even a quick inspection of the structures involved shows that they tend to be larger and more hydrophobic than is common within the standard alphabet. It is possible that the xenoalphabet shown might be capable of forming beta strands, but far less clear whether it could form alpha helices, or beta turns (although it does contain Cys and Ala). 

(Mayer-Bacon et al.)

Having observed that one of the winning sets may not even be capable of forming an alpha helix (a key component of protein structure), they admit the metrics (hydrophobicity, size, charge) used for defining “better” coverage (range + evenness) likely need refinement. Following this admission is a beautiful discussion about refinement of one’s assumptions for the discovery of currently unknown constraints of design in biology. Because consideration of hydrophobicity, size, and charge produced a winning set inconsistent with the real world, an iterative refinement of assumptions is required. This may produce an additional constraint for consideration. (Mayer-Bacon et al.) Accounting for structural aspects like “chemical neighborhood” in proteins and volume of amino acids (Mayer-Bacon and Freeland) will likely reveal the natural set to be even more exceptional than is currently appreciated.

Dr. DeBenedictis seems to suggest the amino acids are “boring” because they only contain five chemical elements out of the 118 on the periodic chart. Actually, hydrogen, carbon, nitrogen, oxygen, and sulfur each have special properties that are critical for the aqueous biochemistry of life. (Atkins) “King” carbon surpasses all the other elements in the number of chemical compounds it can form. (Atkins) The stability of the carbon-carbon bond and tetravalency provide carbon special fitness to form a vast array of organic compounds. Indeed it forms more chemical compounds than all known non-carbon elements combined and to date has around ~10 million known compounds. (Denton; Atkins)

At [10:02] of her TEDx talk, DeBenedictis argues that expanding biology to be more chemically sophisticated can provide solutions to current problems.

Do you want to make proteins that break down plastic bottles? Use [new] amino acids.

That sounds great, but I’d argue that a greater respect for biological design might avoided a plastic crisis in the first place. Molecular compounds might just use that small subset of the periodic chart for important reasons like sustainability and disintegration. (Denton)

In general, molecular compounds [made up of covalently bonded atoms] are the soft face of nature, and ionic compounds [inorganic] are the hard. Few distinctions make this clearer than those between the soft face of the Earth — its rivers, its air, its grass, its forests, all of which are molecular — and the harsh substructures of the landscape, which are largely ionic. This is why the upper triangle of the Eastern Rectangle [in the periodic table] is so important to the existence of life, and why all the rest of the kingdom is so important in the formation of a stable, solid platform.

(Denton)

I don’t disagree that non-canonical amino acids (ncAAs) may have important engineering applications. However, I anticipate (based on past and current examples) that there are advantages to the way nature does things that are currently underappreciated. Because of this I would urge that the role of ncAA’s in designing new products that function outside of living organisms or serve unique roles in living organisms be explored carefully. We want to avoid creating another plastic crisis that is hard to get rid of. Additionally, understanding why the natural amino acids work so well will likely be informative as to what applications different xeno amino acids might be suited for. (Mayer-Bacon et al.) At the heart of the matter is that appreciating biological design considers trade-offs within ecosystems as an important design principle that will prevent production of products that cause more problems than they solve.

So, is the natural set of amino acids a limitation in biology? To date we haven’t found anything better for the aqueous needs of biochemistry. What we have learned suggests that the natural set of amino acids is exceptional. I will leave you with two questions: What evidence do we have that new amino acids will mesh well in living systems and actually improve therapeutics? Which conflicting constraints have been analyzed to confidently make these statements? In my next post, I will discuss DeBenedictis’s example of the INK4a/ARF overlap as poor design.

This post has been updated to reflect a quote incorrectly attributed to Athel Cornish-Bowden and María Luz Cárdenas but was actually written by the German biophysicist Reinhart Heinrich. The quote appeared in a book edited by Cornish-Bowden and Cárdenas, but was in a chapter written by Heinrich.