In 2017, a team led by Rafik Neme published a paper in PLOS Genetics that has since become the standard empirical counter-move in debates about information and origins. They took random RNA sequences, expressed them in living E. coli, and found that roughly one in four increased bacterial growth rate. The conclusion drawn by many who cite it: information can arise from randomness. Natural selection, working on raw chemical variation, can do what intelligence does. The argument for a designing mind, they say, is refuted. This article examines what the study actually shows, what it requires to function, and why the conclusion drawn from it does not survive contact with what information theory actually means.
I. The Argument Being Challenged
The information argument, stated precisely, is not that useful chemical reactions cannot arise without intelligence. It is a narrower and more specific claim: digital code, of the kind found in the genome, cannot arise without intelligence because code requires a prior established convention between encoder and decoder, and no physical process has ever been observed generating such a convention from scratch.
Claude Shannon's 1948 framework defines information as a function of the reduction of uncertainty within a pre-established communication system.[1] A message has information content relative to a receiver that can interpret it. The symbol "A" carries information only because there is already a mapping, a convention, between that symbol and its referent. Without the mapping, the symbol is noise. The genome operates under exactly this logic: the codon AGC specifies serine not by chemical necessity but by convention; a convention maintained by the tRNA machinery, the aminoacyl-tRNA synthetases, and the entire translational apparatus working in coordinated precision.
This is the claim being tested. Not "can random chemicals have effects?" but "can random processes produce specified, conventionally-encoded, functionally coherent information systems from scratch?" These are not the same question. The gap between them is where the debate lives.
II. What Neme et al. 2017 Actually Did
The researchers synthesized libraries of random nucleotide sequences and cloned them into expression vectors in Escherichia coli. They screened approximately 10,000 sequences for effects on bacterial fitness, finding that roughly 25% conferred a detectable growth benefit. From these, they selected three sequences for deeper characterization. Two of the beneficial sequences operated through RNA molecules alone; the RNA itself interacted with existing cellular machinery to produce the effect. One produced a small polypeptide that affected cell behavior. None of the three matched any known gene family. The team concluded that novel, functional sequences could arise from non-coding "junk" space, supporting the concept of de novo gene birth from non-genic sequence.[2]
The study is real, peer-reviewed, and carefully conducted. The data are not in dispute. What is in dispute is what those data establish about the origin of information. There are three layers of analysis required to evaluate this correctly.
III. Layer One: The Machinery Was Already There
Every beneficial effect observed in Neme et al. 2017 was produced inside a fully operational, extraordinarily sophisticated living cell. The random sequences did not build the ribosome. They did not build the tRNA. They did not establish the genetic code. They did not wire the regulatory networks. They were processed by pre-existing machinery of staggering specified complexity.
Imagine inserting a page of randomly typed characters into a functioning library cataloguing system. Occasionally, a random string might accidentally match an existing index entry and redirect a reader to a useful resource. The random page did not write the cataloguing system. It did not create the indexing conventions. It did not establish what "useful" means. It interacted, by accident, with an architecture it did not create.
This is precisely the structure of Neme et al. Random sequences, processed through the pre-existing genetic code and translational machinery, occasionally produced interactions that benefited the cell. The cell's ability to read, translate, and respond to those sequences was not produced by the experiment; it was assumed as a precondition of it.
The information-theoretic claim is about the origin of the code, not its operation once established. Demonstrating that random inputs to an existing system can have beneficial effects says nothing whatsoever about how the system itself came to exist. This is the foundational logical distinction the study's popular citations consistently elide, and it is not a subtle one.
IV. Layer Two: Function Is Not Code
Shannon's information theory makes a precise distinction that underpins the entire debate.[3] Information, in Shannon's framework, is defined relative to a communication system with a source, a channel, and a receiver capable of interpreting the signal according to an established convention. The convention must pre-exist the transmission. Without it, signal and noise are indistinguishable.
A random polypeptide that binds to a membrane protein and alters cell behavior is not demonstrating the origin of a code. It is demonstrating a chemical interaction, a physical event with physical consequences. The same logic applies to RNA secondary structures that happen to interfere with existing cellular processes. These are not symbols specifying anything. They are molecules having effects.
A random peptide binds a protein. A random RNA folds into a structure that interferes with a transcription factor. A pebble in a stream alters the flow. These are physical events. They require no convention. They can be produced by any molecule with the right shape. They produce no symbol, carry no message, and specify nothing.
The codon AGC specifies serine. Not because AGC resembles serine chemically; it doesn't. It specifies serine because the genetic code established that convention, and the translational machinery faithfully implements it. The symbol and referent are arbitrarily related. The convention is the information. No physical process has ever been observed originating such a convention.
Hubert Yockey, a physicist and information theorist who applied Shannon's mathematics rigorously to molecular biology, and who was not an intelligent design proponent, made this point with precision in his 2005 monograph: the genetic code is a mapping of physical triplets to chemical entities that is not determined by any physical or chemical necessity.[4] It is arbitrary in origin and precise in execution. The code, by definition, requires a prior act of convention-setting that no physical process has ever been observed to produce.
V. Layer Three: Selection Cannot Build What Doesn't Yet Exist
The deepest problem with the citation of Neme et al. as a refutation of the information argument is an evolutionary logic error that is rarely examined.
Natural selection operates on heritable variation that already exists. It filters. It amplifies. It preserves what reproduces better. It does not generate new information systems. It works downstream of replication, which means it requires a pre-existing genetic code, a pre-existing translational apparatus, and pre-existing cellular machinery to function at all.
Natural selection requires replication. Replication requires a code (to specify what to copy), a decoder (to read it), and a copier (to implement the instruction). All three must be simultaneously present for selection to begin. Selection cannot build the system it requires to operate; because it cannot operate until the system already exists.
Neme et al. operates entirely inside a cell where all three components already exist and function at extraordinary precision. The experiment demonstrates that random variation, once introduced into an existing information-processing system, can occasionally improve that system's performance. This is unremarkable; and it is not the origin of the system.
The origin question, the only question the information argument addresses, remains completely untouched.
This is not a gap-of-the-gaps argument. It is a logical argument about the conditions necessary for selection to operate at all. The philosopher of biology Elliott Sober has argued that Darwinian selection has genuine explanatory power for the diversification of life; but that it presupposes, and therefore cannot explain, the origin of the replication system itself.[5] The most honest origin-of-life researchers have acknowledged this for decades.
VI. The Three Studies and What They Actually Establish
The challenge often cites three studies as a package. Each deserves its own accounting.
Neme et al. 2017: Examined above. Shows random sequences can interact beneficially with pre-existing cellular machinery. Does not address the origin of that machinery or the code it uses. Does not produce a new symbol-referent convention. Fully consistent with the information argument.
The broader de novo gene birth literature: Several studies (Ohno 1970; McLysaght and Hurst 2016; Vakirlis et al. 2020) document cases where non-coding genomic regions acquired function over evolutionary time.[6] The critical detail universally overlooked in popular citation: in every documented case, the "new" gene emerged within an existing genome with an existing code, an existing cellular environment, and existing regulatory scaffolding that could detect and preserve beneficial variants. The code was not built. The code was used.
Functional sequence space studies: Research by Douglas Axe (2004) and Ann Gauger et al. estimated the ratio of functional to non-functional protein folds in sequence space.[7] Axe's estimate: approximately one in 1077 sequences of typical protein length (150 amino acids) folds into a stable, functional structure. Neme et al.'s 25% figure applies to any detectable cellular effect in a living system already optimized for responsiveness, a vastly lower bar than coding for a stable, functional, heritable protein fold. These findings are not in tension. They are measuring different things.
VII. The Coherence Problem: What Studies Like This Cannot Address
Even granting every result in Neme et al. at full face value, the study addresses isolated beneficial interactions: single sequences, single effects, single fitness benefits in a controlled environment. The genome is not a collection of isolated beneficial interactions. It is a schematic system: integrated, interdependent, multilayered, and coherent across its entire length.
The human genome encodes approximately 20,000 protein-coding genes, regulated by hundreds of thousands of regulatory elements, organized in three-dimensional chromatin architecture, corrected by multiple overlapping repair systems, and interpreted by a translational apparatus that is itself specified by a subset of the same genome. No part of this system functions independently of the rest. No incremental random improvement to one component produces the preconditions for the next, because the system's function depends on simultaneous precision across all components.
Stephen Meyer's argument in Signature in the Cell, which engages the origin-of-life literature in technical detail, focuses precisely on this point:[8] the genome does not resemble a collection of independently selected parts. It resembles a designed schematic. The difference matters because schematics are not assembled incrementally by blind processes; their coherence is a precondition of their function, not a product of it.
Consider the difference between a collection of metal parts and a working watch movement. Natural selection can polish a gear that already exists. It can discard a gear that doesn't fit. It cannot design the schematic that specifies which gears are needed, at what tolerances, in which order, for a system that doesn't yet exist.
The Neme 2017 finding is a polished gear, a random part that happened to interact usefully with an existing mechanism. The origin-of-life question is: where did the watch come from? The gear does not answer this. Neither does finding ten thousand gears.
VIII. The Honest Position on Both Sides
The information argument is not claiming that random chemistry is irrelevant to biology, or that beneficial mutations do not occur, or that gene families cannot expand. These are not the claims. The information argument claims something precise: the origin of the genetic code, the arbitrary, conventionally-established mapping between nucleotide triplets and amino acids, and the machinery that implements that mapping; has not been, and by the logic of information theory cannot be, explained by any purely physical process observed to date.
Wherever this claim has been honestly examined, it holds. The best origin-of-life researchers in the world, not creationists, not intelligent design theorists, but mainstream biochemists; have acknowledged that the origin of the code remains the hardest unsolved problem in all of science.
That quote is from Eugene Koonin, one of the leading evolutionary biologists in the world, a prolific author, a committed evolutionist, and a scientist who is rigorously honest about what the data show. He is not arguing for intelligent design. He is simply acknowledging that no mechanism for the origin of the code has been found; and that the universality of the code suggests it happened exactly once, under conditions we have no way to reconstruct.
Against this, the citation of Neme et al. as a refutation is an overreach that the paper itself does not support. It is a category error: using evidence about beneficial mutation within an existing coded system to dismiss questions about the origin of the coded system itself. These are not the same question. Honesty about the distinction is not creationist special pleading. It is precisely what the best evolutionary biologists themselves require of the field.
IX. The Site's Position
This site does not claim omniscience about origins. The evidence for the information argument is strong, strong enough to constitute a genuine defeater for purely materialist accounts of the genome's origin, in the judgment of serious philosophers of science. But it is presented here as an argument, not a proof.
What we will not do is accept the misuse of Neme et al. as a conversation-ender when it is not. A study demonstrating that random sequences interact usefully with existing cellular machinery does not explain the origin of that machinery. It does not produce a new code. It does not bridge the gap between chemistry and convention. The gap is real. It has been acknowledged by every honest researcher who has spent a career trying to close it. That acknowledgment deserves to be heard.
A Convention Requires a Convenor
Every arbitrary symbol-referent mapping in human history was established by a mind. Every convention was set by an agent. The genetic code, the most precise and universal symbolic convention ever discovered, operates on the same logical structure. No physical process selects which codon means which amino acid because no physical necessity determines it. The code is arbitrary. Arbitrary codes require agreement. Agreement requires minds.
If the code was set, and the universality of the genetic code suggests it was set once, precisely, and maintained without variation across all life on Earth, then it was set by something. The question that remains is not whether there was a Coder. The question is who. And the claim of this site is that the historical record supplies an answer that demands serious engagement, not dismissal.
The Information Theory Argument in Full →The following sources constitute the primary intellectual foundations for preparing a rigorous response to the information-from-randomness challenge. Read the primary studies before citing them in either direction.
- Neme, R., Amador, C., Yildirim, B., McConnell, E., Bhatt, S., & Tautz, D. (2017). "Random sequences are an abundant source of bioactive RNAs or peptides." PLOS Genetics 13(6): e1006839. The study being examined in this article. Read it before citing it. The paper's claims are about beneficial interactions with existing cellular systems, not about the origin of those systems. The authors do not claim to have addressed the origin of the genetic code. PubMed · Full text ↗
- Shannon, C.E. (1948). "A Mathematical Theory of Communication." Bell System Technical Journal 27(3), 379–423. The foundational document. Shannon's definition of information is precise and requires a pre-established communication system. He was solving an engineering problem, not writing metaphysics. His framework is what both sides of this debate must engage, and what the popular citation of Neme et al. consistently fails to engage. Read source (PDF) ↗
- Koonin, E.V. & Novozhilov, A.S. (2009). "Origin and evolution of the genetic code: the universal enigma." IUBMB Life 61(2), 99–111. The single most useful source to cite in this debate. Koonin is a leading evolutionary biologist who acknowledges that the origin of the genetic code is "the most difficult problem in evolutionary biology" and remains unsolved. Read it, understand it, cite it, not as a creationist reference, but as the honest acknowledgment of an established field leader. Search this source ↗
- Meyer, S.C. (2009). Signature in the Cell: DNA and the Evidence for Intelligent Design. HarperOne. The most rigorous philosophical treatment of the information argument. Meyer's doctorate from Cambridge in history and philosophy of science gives this work a precision that popular ID writing lacks. Chapter 10 directly addresses the "chemical necessity" and "chance" hypotheses for the origin of the code. Engage the strongest version of the argument before dismissing it. Search this source ↗
- Yockey, H.P. (2005). Information Theory, Evolution, and the Origin of Life. Cambridge University Press. Yockey was a physicist who applied Shannon's mathematics to molecular biology with rigor and without creationist agenda. His conclusion: the genetic code is informationally irreducible to chemistry. The sequence space analysis in this book is the technical foundation for understanding why Neme et al.'s 25% figure and Axe's 10⁻⁷⁷ figure are measuring entirely different things. Search this source ↗
- Axe, D.D. (2004). "Estimating the Prevalence of Protein Sequences Adopting Functional Enzyme Folds." Journal of Molecular Biology 341(5), 1295–1315. The most important empirical study on functional sequence space. Axe estimated that approximately 1 in 10⁷⁷ sequences of protein length folds into a stable, functional enzyme. This figure is frequently dismissed. Read the paper itself and the peer response to understand why it has not been successfully refuted on methodological grounds. Search this source ↗
- Orgel, L.E. (1973). The Origins of Life: Molecules and Natural Selection. Wiley. Orgel spent his career at the Salk Institute working on origin-of-life chemistry. His honest assessment of the code-origin problem, "the most baffling aspect of the origin of life," is quoted in this article. He was not a creationist. He was a rigorous biochemist who called the problem what it is. Search this source ↗
- Sober, E. (2008). Evidence and Evolution: The Logic of Darwinism. Cambridge University Press. Sober is a philosopher of science and committed Darwinist who argues that natural selection has genuine explanatory power for diversity and adaptation — but who is precise about what it can and cannot explain. His analysis of the preconditions for selection to operate is useful for understanding Layer Three of this article's argument. Search this source ↗
- McLysaght, A. & Hurst, L.D. (2016). "Open questions in the study of de novo genes: what, how and why." Nature Reviews Genetics 17, 567–578. A thorough review of the de novo gene birth literature. Relevant for understanding what the evidence for random-sequence-to-function actually shows — all documented cases arise within existing genomes, using existing codes, within existing cellular environments. The article is honest about the difficulty of establishing true de novo origin. Search this source ↗
- Wilder-Smith, A.E. (1981). The Natural Sciences Know Nothing of Evolution. Master Books. Wilder-Smith had three earned doctorates in natural science and was among the first to apply information-theoretic thinking to biology in a rigorous way. His argument about symbolic language in the genome — that the symbol is not its referent, and therefore codes require prior convention — predates Meyer's work by decades and remains under-cited in the academic literature. Search this source ↗