Meyer and Nelson on a Failed Explanation for the Origin of the Genetic Code
Ann Gauger has already drawn our attention to the new paper, published just last week, in the journal BIO-Complexity. Authored by Discovery Institute's Stephen Meyer and Paul Nelson, the paper is concerned with the question of the origin of the genetic code, and seeks to evaluate the efficacy of the so-called Direct RNA Templating (DRT) hypothesis as an explanation for its origin.
First formulated and proposed by Michael Yarus in 1998, the DRT scenario has enjoyed support from such figures as Ma Wentao, and I.S. Dunn. Before we turn our attention to the evaluation of the merits of the idea, a little background and contextualization will be helpful.
What Needs To Be Explained?
First, let us consider what the scenario is being invoked to explain. The key points of the mRNA-to-protein translation system (summarized in the Meyer and Nelson paper) can be visualized with the aid of the following animation:
What is Direct RNA Templating?
For reasons that I have reviewed and critiqued previously (here and here), the RNA world scenario enjoys a broad base of support from contemporary origin-of-life theorists. The DRT model takes this a step further. It argues that the idea is not only sufficient to circumvent the chicken-and-egg conundrum pertinent to the relationship between DNA and proteins, but that it is also adequate to explain the origin of the genetic code itself. This is, naturally, of particular interest to proponents of intelligent design. ID takes the existence of such digitally encoded information as a central argument for explaining life, at least in part, as the product of an intelligent cause. If this complex specified information (CSI) could be adequately accounted for in materialist terms, that would render the design postulate redundant.
So what, exactly, does DRT suggest? Yarus and his supporters maintain that specific peptides were first made by ordering carboxyl-activated amino acids using amino acid sites in an RNA template. Within the sites, it is proposed, are subsequences which will be selected as the translation system evolves into what it is today. RNA polymers with particular sequences of bases would attract the specific amino acid with which they are currently associated. Ordered peptides themselves are envisioned as being of value in the RNA-based biosphere such that there was a selection for the origination, and subsequent incremental enhancement, of protein-synthesizing (translation) machinery.
Ma Wentao describes the DRT scenario as "attractive because it should be the simplest mechanism for RNA to synthesize peptides, thus very likely to have been adopted initially in the RNA world. Then, how this mechanism could develop into a proto-translation system mechanism is an interesting problem."
Offered Experimental Support
In 1988, Michael Yarus published a paper in the journal Science. In this study, as relayed by Meyer and Nelson,
...he discovered a differential bonding affinity between the amino acid arginine and RNA bases at the active site in the group I intron of Tetrahymena, a ciliated protozoan. He found that arginine inhibits the self-splicing reaction of the group I intron by preferentially binding to sequences containing nucleotides corresponding to arginine codons (AGA, CGA, and AGG). These data led Yarus et al. to speculate that the group I intron represented a molecular fossil--showing the specific binding of amino acids directly by RNA--which he claimed 'developed from an ancient RNA codon-amino acid interaction'.
In a later (2009) paper in the Journal of Molecular Evolution, Yarus and his colleagues reported that, for six out of the eight amino acids they studied (the only exceptions being glutamine and leucine), codons were "unexpectedly frequent in cognate RNA-amino acid binding sites." They suggest that "there was likely a stereochemical era during the evolution of the genetic code, relying on chemical interactions between amino acids and the tertiary structures of RNA binding sites."
Is this the end of ID, as some commentators have suggested? Can the genetic code be constructed by virtue of undirected stereochemistry? And, if so, does this serve as a potent counter-example to the purported uniform causal relationship between information and intelligence? Meyer and Nelson don't think so. Now, let's turn to the reasons why.
Selective Use of Data
Meyer and Nelson begin their critique by highlighting the apparently cherry-picked nature of the data set analyzed by Yarus and his colleagues, noting that the choice of which aptamers (i.e., RNA molecules that bind to a specific target molecule) one analyzes can have a significant impact on the statistical validity of what he is attempting to demonstrate. In other words, such a pattern should not be forced or artificially imposed upon the data set.
Meyer and Nelson ask,
Have Yarus et al. introduced such biases into their statistical analysis? The answer appears to be a troubling 'Yes,' as a careful analysis of their 2009 review article makes clear.Yet Meyer and Nelson highlight further evidence that Yarus and his colleagues set aside data that was inconsistent with their hypothesis.They continue,
Recall that SELEX methods may capture a diversity of RNA sequences "that perform the same task." For example, when isolating tryptophan-binding RNAs, Yarus et al. found RNA sequences with a conserved region, which they dubbed the CYA Trp motif. But they also found 19 unique sequences that, while binding tryptophan, lacked the CYA motif. Yet Yarus et al. failed to consider these sequences in their analysis. [see Majerfeld and Yarus, 2005]
The SELEX trawl captured several RNA sequences that bind tryptophan. Therefore, to avoid bias, all of these sequences should be analyzed statistically -- not simply the motifs that look interesting on the stereochemical hypothesis (i.e., sequences exhibiting a disproportionate representation of code-relevant triplets). Otherwise, the screening criteria may artificially amplify the signal the investigators purport to have found -- rather like catching both salmon and mackerel, throwing away the mackerel, and then claiming that the trawl caught only salmon.
Additional evidence...can be found when one examines amino acids that don't appear in the list of eight above. Consider valine, for instance. One might think that Yarus et al. had yet to investigate RNA binding affinities for valine, but in fact, they did.Thus, as Meyer and Nelson note, "the claim of Yarus et al. 2009, that the valine RNA aptamer results were too poorly characterized to allow their inclusion in the statistical analysis, appears to be contradicted by their earlier publications."
So why doesn't valine figure in the 2009 statistical analysis or in Yarus's book on the subject [You can purchase this book here]? They did not find code-relevant triplets in the binding site of the valine aptamer.
Incorrect Null Hypotheses
Meyer and Nelson also highlight a different (but related) problem concerned with the election of erroneous null hypotheses. They note that Yarus et al. "tested for a higher concentration of cognate codons in the amino-acid binding sites of their aptamers, as opposed to the non-amino acid binding nucleotides of the aptamers. But this would be the correct null hypothesis only if Yarus et al. had examined all relevant RNA sequences (aptamers)."
Meyer and Nelson continue,
The correct null hypothesis asks whether non-cognate triplets are found as often as cognate triplets in the binding sites of all aptamers for a given amino acid. However, because Yarus et al. evidence little curiosity about those unique aptamers that bound amino acids, yet lacked conserved sequence motifs, it is impossible to use the correct null hypothesis. The other sequences have already been tossed back into the ocean. The null hypothesis Yarus et al. actually employed, therefore, asks only about the frequency of cognate triplets in the binding sites of the aptamers that they selected for analysis -- which looks exactly like the sort of illegitimate statistical bias Ellington et al. [Link] described as "imposed...by man."
Meyer and Nelson go on to mention the criticism made by Koonin and Novozhilov (2009) that,
...the affinities are rather weak, so that even the conclusions on their reality hinge on the adopted statistical models. Even more disturbing, for different amino acids, the aptamers show enrichment for either codon or anticodon sequence or even for both, a lack of coherence that is hard to reconcile with these interactions being the physical basis of the code.In other words, both the codon and the anticodons are found in multiple aptamer binding sites. But there is no feasible scenario that would facilitate a simultaneous role of the codon and the anticodon in translation.
Implausible Geometry for the Primordial RNA Template
The next part of the paper discusses the universally conserved structures of modern ribosomes and the peptidyl transferase center (the component of the ribosome in which peptide bonds are synthesized between amino acids by the enzyme peptidyl transferase). This structure has a very specific three-dimensional shape, which facilitates the specific positioning and movement of the amino-acid carrying CCA stems of adjacent tRNAs to facilitate the the synthesis of peptide bonds.
The problem is that, in the DRT scenario, such precision does not exist. Meyer and Nelson note,
...even if an ensemble of RNA aptamers aligned in close proximity to one another, and even if they did so in a way that would in theory specify an amino acid with biological relevance...no evidence shows that amino acids thus carried by the RNA aptamers would form peptide bonds, especially in any realistic prebiotic setting.
Meyer and Nelson identify another related problem with Yarus's model:
In extant cells, the tRNAs that hold amino acids in place for peptide bond formation do so using covalent bonds. These strong chemical attachments enable the tRNA to present the amino acid at a distance from the main body of the tRNA molecule, to prevent any steric hindrance to peptide bond formation. The DRT model RNA aptamers, however, bind amino acids using weaker non-covalent associations. As a result, the RNA aptamers have to make more extensive contact with their amino acid ligands.Unsupported assumptions about the pre-biotic availability of amino acids.
This raises the possibility that the RNA aptamers will either partially, or completely, envelope the amino acids to which they are bound, or that they will otherwise introduce steric hinderance to peptide bond formation. Recognizing this problem, Yarus et. al have carefully engineered their aptamers to ensure that they attach to the side groups of their corresponding amino acids, rather than only to the α-amino and αcarboxyl groups, where peptide bonds form. This engineering clearly represents intelligent design, and thus does not simulate an undirected stereochemical origin of the genetic code, but rather its opposite.
On page 7 of their paper, Meyer and Nelson turn their attention to Yarus's unjustified assumptions about the pre-biotic availability of amino acids, pointing out that the side chains of the amino acids, tested by Yarus and his colleagues, are fairly complex. The amino acids that are missing are those with the simpler side chains (e.g., alanine, serine and glycine). But the biosynthetic pathway leading to many of the more complicated amino acids (e.g., tryptophan) are accordingly far from trivial, involving many enzymes and functionally-integrated biosynthetic pathways.
The problem here, for Yarus et al., is that a key prediction of the model is that the very simplest (i.e., the easiest to make) of the amino acids should exhibit the stereochemical affinities. So why were they left out of the analysis? This is especially curious given that valine failed entirely to correlate with the model!
Chicken and Egg
Yarus's model also raises a significant chicken-and-egg paradox. Meyer and Nelson explain:
Because those biosynthetic pathways involve many enzymes, extant cells would require a pre-existing translation system in order to make them. Since attempts to explain the origin of the genetic code are also attempts to explain the origin of the translation system (indeed, there can be no translation without a code), Yarus et al.'s findings raise an acute chicken and egg problem. Which came first, the aptamer-amino acid affinities that Yarus et al. propose as the basis of the code and translation system, or the translation system that would have been necessary to produce those amino acids (and, thus aptamer-amino acid affinities) in the first place?
Summary and Conclusion
Meyer and Nelson highlight a plethora of problems in Yarus et al.'s experimental methodology and theoretical model. The methods of selecting amino-acid-binding RNA sequences failed to take into account aptamers that failed to conform to the predictions of the model. Moreover, the reported results exhibited a 79% failure rate, which calls into question the validity of the supposedly "correct" results. The model is also fatally flawed for reasons such as the chicken-and-egg conundrum, the implausible geometry for the primordial RNA template and the unjustified assumptions regarding the availability of prebiotic amino acids.
Meyer and Nelson wrap up by drawing attention to the positive argument for ID based upon uniform cause-and-effect relations: codes and digital information are categories of effects uniformly associated with intelligent causes. Indeed, to the extent that Yarus and others have succeeded in establishing affinities between codons and amino acids, they did so only as a direct consequence of their own intelligent manipulation and intervention. But if we require intelligent input to accomplish such a feat under laboratory conditions, how can we expect the blind and impersonal forces of nature to be up to the task?