Reviewer 2: David Krakauer (Santa Fe Institute)
As this paper is somewhat unusual, consisting more of a philosophical contribution at this stage than a scientific one, my review will be concordantly, somewhat unusual. I should declare at once that I do not share Professor Koonin's perspective on the early origin of life and the role of what is sometimes called " anthropic" reasoning or the self-selection principle. Anthropic reasoning has proved to be of some utility in string theory where the superabundance of meta-stable low energy vacua (or distinct universes with distinct parameters) realized through inflationary mechanisms, has proved a challenge to those projects seeking to derive our universe from first principles. The weak anthropic principle seeks to determine what we can expect to observe by the conditions neccessary for our presence. The principle first suggested by Carter is a perfectly scientific statement, and somewhat surprisingly has value in allowing us to apply Bayesian inference to cosmological phenomena. Carter was also the first to apply anthropic principles to the origin of life in 1983, where it was argued that many steps in the early evolution of life might have been fairly improbable, an argument pursued at greater length in this contribution. My principal objection throughout this review will be that the anthropic principle is only scientific in so far as it can do some work, in allowing us to calculate different states of order. Without the calculation it remains an interesting metaphysical insight.
Author's response: I do not strongly disagree except that I believe that the demand for "calculation" is unnecessarily narrow. To be scientifically useful, a principle (concept, hypothesis etc) must offer testable predictions, quantitative or qualitative. Both kinds of predictions are given in this manuscript, qualitative ones in the body of the paper, and quantitative ones in the Appendix. Granted, the latter are toy calculations but calculations nevertheless. Thus, by this simple criterion, this work is "physics" not metaphysics, even as metaphysical implications abound.
The Koonin Hypothesis (KH for short) is clearly stated on p12 of the manuscript, "The core elements of the translation system, namely a RNA -only ribosome and the specific adaptors for, at least, a subset of the 20 modern protein amino acids emerged by chance and were anthropically selected." In other words, one of the core structures upon which all of life depends arose in a series of highly improbable steps. Following the theory of large deviations, replication and translation emerged as an exponentially unlikely event in proportion to its deviation from a suitable equilibrium ground state. This we might call an "ahistorical" theory of proto-biology.
Author's response: I have no objection to Krakauer's statement that the quoted phrase formulates KH; let it be so. However, to completely disambiguate the issue, I should reiterate that KH is just a strong, extreme version of a more general framework that, in the same tongue-in-cheek style, I will denote KC (Koonin Concept). The KC can be stated thus: biological evolution started when the minimal complexity required for that was reached through processes governed by anthropic selection. Further, I think that defining KC/KH as "ahistorical" is somewhat misleading. There is history involved, for sure, just not biological evolution/selection. Speaking of these concepts as "non-selectionist" or "non-adaptationist" would be more accurate.
The history of science, is by one reading, interpreted as a catalogue of conceptual bottlenecks overcome by synthesizing unlikely concepts. The characteristics of a bottleneck are: an inability to make progress based on contemporary theory and data, a general impetus to search for radically new ideas, and a correlated tendency to default to extra-scientific modes of explanation based on putative forces and miracles. The general theory of relativity, quantum mechanics and natural selection all solved hard problems by combining hitherto unrelated concepts – non-Euclidean geometry and gravity, probability and mechanics, density regulation with environmental selection. Interestingly, all three theories have their non-scientific resonances, "everything is relative", "everything is subjective", and "everything was created" descending from the bottleneck period. The origin of life will require just such juxtapositions and I find that resorting to extreme fluctuation reasoning, such as in the KH, somewhat unsatisfactory
I think that the KH falls short of being a scientific statement as it invokes events with no well defined probability measure, and in addition, eschews identifying mechanisms.
First, with regard to probability spaces, recall that any probability space includes a sample space of events (which we can think of as the power set), a set of outcomes or events (non-empty subsets) the sigma algebras, and a probability measure that assigns a real value in the interval 0 to 1 to the events. This value is analogous to a volume in a state space, and tells us how likely it is that an event takes place. Now not all events in the power set have a Lebesgue measure, in other words, there are those to which we can not assign a volume. This is not as strange as it seems, as the rational numbers have exactly this property. In the KH we are provided with no means of calculating the probability distribution over biophylic universes, just the statement that any universe is possible. The anthropic principle does the work of extracting our universe by a posteriori declaring that we are in one. This is logically correct but operationally without value. It has been stated that the whole history (and future) of the world exists in the decimal expansion of pie (assuming that it is random). But the probability that we would find this history is indeterminate. This is what I worry about in the KH, that there is no way of assigning a measure to the replication and translation machinery in the multiverse.
In this paper we have neither any mechanism or constraints for the translation system, nor prior observations on which to base our probability measures. We have only a conjectured sample space, and no systematic mechanism of coarse graining the relevant measures. This leads to a rather strange outcome where we might as well assert that all observed biological order emerged in one step, including the complete evolutionary history of life. This is equally as possible as the emergence of an RNA polymer, and eye or an atom. Notice that these are all equally possible, namely certain in an infinite multi-verse, as there is no conceivable way of deciding whether they are all equally probable without first writing down an appropriate equilibrium theory for the universal ground state, by making use of relevant cosmological data etc. in our local space.
Author's response: A probability measure for any events is readily defined in the MWO multiverse – in fact, the definition of probability is simpler and more natural than it is in a finite world inasmuch as probability and frequency become one and the same (see the new Table 1 I made per Bapteste's suggestion).
Now let's consider mechanisms. Mechanisms represent our most fundamental intuitions about causality; mechanisms posit relationships between two or more observables based on a theory of interactions. Mechanistic theories are always provisional, but their utility can be inferred through their ability to provide useful predictions out of sample. Mechanisms offer an alternative to observation, by interpolating and extrapolating from a finite body of data. Statistical associations are useful but they are limited to a single history of observation. Mechanisms rise above history and are capable of describing an ensemble. Darwin's theory derives its power not from its ability to explain the bill of a finch, the shell of a snail or the leaves of the mimosa plant, but because it can account in part for all of these, and indeed any, adaptive system in nature, on earth or elsewhere.
In this paper we are provided with no constructive algorithm for arriving at a probability measure for translation or any natural mechanism by which it might come about. The reason why the KH stops short of invoking spontaneous creation for all levels or organization, is that it is confident that we have a suitable theory – Darwin's theory – once a certain threshold of biology has been reached. But this hybrid solution is rather strange as it reveals a possible inconsistency. If we are faced with a mechanism, the KH suggests that we should select this above a fluctuation-based hypothesis. This is prescribed even when the KH is fully able to account for any level of organization compatible with the laws of physics. In the appendix to the paper, an informal calculation of the chance emergence of RNA polymers is provided, whose accuracy I can not judge within a few orders of magnitude. A very low probability is adduced and then used to support an infinite multiverse (O-ring) anthropic selection argument. But as I have stated above, an infinite model allows for infinite complexity at any level, and so the improbability of the polymer is not really crucial. On the basis of the KH are we really able to declare that any level of organization above replication and translation represents a sufficiently large fluctuation to be radically less probable, and therefore in need of natural selection? This is an interesting metaphysical idea, but not yet a scientific one.
So what's going on here? I think that we can identify two primary motivations for this paper: (1) a perhaps somewhat restricted application of scientific reasoning, and (2) an excessive dependence on natural selection.
I believe we have reached a stage of knowledge where we need to be pluralistic about the "scientific method". Most of science has the following features: (1)testable or refutable, (2) minimal, (3) extendable or generalizable, and (4) probable. Not all theories in science possess all of these properties, but all of them come up in debate in one form or another. The first or refutability is the Popperian gold standard. It defines the demarcation criterion of science from pseudoscience, distinguishing assertions that can garner consensus from matters of taste and prejudice. Theories that are not refutable in principle do not belong in science. The second is minimality and is closely related to formalization in mathematics. Science seeks the shortest description length for regularities in data. We can always overfit data and this leads to poor prediction out of sample. Science values minimality both because it aids comprehension and because it aids prediction. Good scientific theories are generalizable – Newtonian mechanics works equally well for apples and for planetary masses. Newton described this propertry as self-conformable. Darwins theory has the same property, it works for bacteria and blue whales. The probable criterion relates to inter-theoretic reduction or compatibility. Darwin made extensive use of artificial selection to explain natural selection. Darwin was building a prior in the space of concepts, in order to make natural selection intelligible. Theories that have no prior strike us as somewhat unconvincing.
We can now ask how well the KH performs by comparing it against the template of constituents of the loosely specified scientific method. The hypothesis fails the refutation test prima facie. The KH declares that all organizational levels are possible. Even if we were to discover a selective theory or a physical self organizing theory for replication and translation, the KH could still be true. Nothing can render it false. Presumably Koonin would no longer support the anthropic theory, not because it is any less possible, but because it lacks many of the ingredients that we have come to value in our scientific ideas. With regard to minimality, the theory is certainly minimal as it requires only the laws of physics, but here minimality does not serve the goal of prediction. Moreover, the theory is not at all extendable as every realization arises through a unique series of events. And we are not provided with a calculation for the probability of translation spontaneously emerging and fixing. Hence the theory does not fit well into what we normally think of as science.
It would seem that the primary motivation for the KH is the absence of a compelling Darwinian account for translation. But the Darwinian theory is only one of many alternative theories that seek to explain ordered states in nature. Before natural selection there existed physics, chemistry and geology. There is a growing interest in the possible role that abiotic dynamics have played in the origin of primary metabolism and primordial replicating systems. Indeed some of my colleagues have recently proposed that: "the reverse citric acid cycle is statistically favored among competing redox relaxation pathways under early-earth conditions and that this feature drove its emergence and also accounts for its evolutionary robustness and universality. The ability to enhance the rate of core reactions creates an energetic basis for selection of subsequent layers of biological complexity." (Smith and Morowitz 2004). This is an alternative to a selective account as under pre-enzymatic conditions, all reactions that relax the free energy of more stable, small inorganic molecules are impeded. Thus the essential constituents of primary metabolism emerge as a reduction sequence in a physical process of free energy minimization. This is a theory that requires a great deal of further research, both theoretical and empirical, but it conforms to the more traditional standards that we have come to require of a scientific approach.
Science seeks to understand the probable, and moreover assign some measure to this concept. There are many ideas that are possible-yet-untestable, but I see no reason to opt for these approaches when much traditional science remains to be done on the early origin of life, some of which suggests that life is far more probable than we have supposed. In an infinite multiverse all life is inevitable, and so we would have little reason to favor one parsimonious theory over another.
Author's response: I agree with Krakauer that (as long as one is interested in a definition of science) a pluralistic definition would be required. However, I do not think this is the place to discuss such a definition in general terms. Suffice it to say that Popperian falsifiability of KH is strongly emphasized in this paper. The falsifiability of the general framework of KC is a more touchy matter as almost always is the case with general concepts, However, a full success of the search for an overarching selective principle that would precede and override replication of genetic material, might falsify KC in full. The work of Smith and Morowitz (, which is mentioned by Krakauer, (Smith, E., Morowitz, H.J. Universality in intermediary metabolism. Proc Natl Acad Sci U S A. 2004;101:13168–73) belongs in this area as well as even more detailed proposals of Pross (e.g., Pross, A. The driving force for life's emergence: kinetic and thermodynamic considerations. J Theor Biol. 2003; 220:393–406). Certainly, this is a legitimate and, potentially, interesting research direction. However, I agree with Krakauer that a lot of work is required to take these studies to a stage where it can be positively claimed that selection, in a meaningful sense, occurred before the advent of replication of digital carriers of genetic information. When and if that happens, KC could become obsolete in its entirety. Until then, however, I believe that a lot of caution is due. Indeed, without implying any accusation, I would note that there seems to be a rather slippery slope here: a non-critical insistence on selection prior to replication might come suspiciously close to some sort of "animism" (sensu Monod).
The history of science, is by one reading, interpreted as a catalogue of conceptual bottlenecks overcome by synthesizing unlikely concepts. The characteristics of a bottleneck are: an inability to make progress based on contemporary theory and data, a general impetus to search for radically new ideas, and a correlated tendency to default to extra-scientific modes of explanation based on putative forces and miracles. The general theory of relativity, quantum mechanics and natural selection all solved hard problems by combining hitherto unrelated concepts – non-Euclidean geometry and gravity, probability and mechanics, density regulation with environmental selection. Interestingly, all three theories have their non-scientific resonances, "everything is relative", "everything is subjective", and "everything was created" descending from the bottleneck period. The origin of life will require just such juxtapositions and has been generating just such non-scientific propositions.
Author's response: I further enthusiastically agree with Krakauer that a productive study of the origins of life requires juxtaposition of "hitherto unrelated concepts". Indeed, this paper is an attempt, however imperfect one, to do just that. I do not believe, however, that the comparison of this work to the proverbially ridiculous "everything is relative" serves any purpose.
Answers to all your Biology Questions
