There Is No "Genetic Architecture of the Human Face"

What I call the Great DNA Myth is a false teaching that continues to be spread by innumerable parties in the world of biology, even though there are very many other authorities in that same world who are telling us the teaching is false.  The Great DNA Myth is the myth that inside DNA is some blueprint or recipe that specifies how to make a human body.  

There are various ways in which this false idea is stated, all equally false:

• Someone may describe DNA or the genome as a blueprint for an organism.
• Someone may describe DNA or the genome as a recipe for making an organism.
• Someone may describe DNA or the genome as a program for building an organism.
• Someone may claim that DNA or genomes specify the anatomy of an organism. 
• Someone may claim that genotypes (the DNA in organisms) specify phenotypes (the observable characteristics of an organism).
• Someone may claim that genotypes (the DNA in organisms) "map"  phenotypes (the observable characteristics of an organism) or "map to" phenotypes.
• Someone may claim that DNA contains "all the instructions needed to make an organism."
• Someone may claim that there is a "genetic architecture" for an organism's body or some fraction of that body. 
• Using a little equation, someone may claim that a "genotype plus the environment equals the phenotype," a formulation as false  as the preceding statements, since we know of nothing in the environment that would cause phenotypes to arise from genotypes that do not specify such phenotypes. 

Weaker formulations of this false idea include claims that DNA is "life's instruction book" or "the key to life" or "the book of life" or "the secret of life." While such rather vague assertions are not as explicitly false as the statements in the bullet list above, such formulations are equally misleading, as they insinuate the false claims in such a bullet list. 

There is no truth to the claim that DNA is a specification for anatomy.  DNA merely specifies low-level chemical information such as which sequences of amino acids make up polypeptide chains that are the starting points of protein molecules.  Many biology authorities (some of which I quote below) have confessed this reality that DNA does not specify anatomy. But the "useful stooge" that is the Great DNA Myth continues to be taught or suggested in the literature of biology by many other people.  So now we have a very strange situation that might be described like this: biology's left hand is writing one thing, and biology's right hand is writing the opposite.  

A recent scientific paper was another example of biology's left hand writing claims that biology's right hand is denouncing as false. The paper had the very misleading title "Insights into the Genetic Architecture of the Human Face." There is no such thing as a genetic architecture of the human face. Genes do not specify the human face or any part of the face or even the structure of any cell that is part of the face. 

The study  found no genes specifying any particular part of the human face.  The study merely found very weak statistical association between some genes and various aspects of the human face. The study claims to have found  "203 genomic regions associated with normal-range facial variation." 

How were such genes found? The technique used by the "Insights into the Genetic Architecture of the Human Face" study involved a  shady trick very much like the shady trick used by DeepMind's AlphaFold2 software in analyzing protein folding.  Trying to help explain how 3D protein shapes arise from mere one-dimensional sequences of amino acids, the AlphaFold2 software used the shady trick of scanning vast protein databases storing the amino acid sequences and 3D protein shapes of countless proteins, a database derived from studying very many different organisms. Why can we call this a shady trick? It's because inside a human body there is virtually none of the information in such databases. Humans do not have inside them some database of the 3D proteins shapes of a vast number of non-human organisms.  The AlphaFold2 software should have followed a rule of "use only information available inside a human being."  Instead it relied on a vast amount of biological information existing only outside of a human being. Since such a technique involves information not available in a human being, it should not be used by any software that claims to help explain how human genes can give rise to the 3D protein shapes found in humans. 

It's a very similar story with the "Insights into the Genetic Architecture of the Human Face" study. The study relied on databases containing scans of very many human faces, along with genomic data (DNA data) for the same individuals, and also data on their gender and ethnicity. It is easy to see how the type of complicated computer algorithms used by the study might have been able to come up with some weak correlations between facial structure and genetic structure. What is going behind the scenes in studies like this is probably something like this:

(1) The algorithms identify genetic variations (called SNP's) associated with a female gender or a male gender, and look for any weak correlation between such variations and facial variations (very easy, because male faces tend to differ from female faces).

(1) The algorithms identify genetic variations (called SNP's) associated with particular ethnic groups, and look for any weak correlations between such variations with facial variations. This is very easy, because certain ethnic groups tend to have certain facial characteristics. 

In the study there are a few lines suggesting a bit of ethnic filtering to limit the ethnic diversity of the people analyzed. We are told that "recruitment was limited to individuals aged 3 to 40 years old and of self-reported European ancestry." But such a technique would not have done much to limit the ethnic diversity of the data analyzed.  For example, a person with Asiatic facial features or African facial features may have a mother from one region of the globe and a father from another region of the globe. If such a person is asked to specify his region of ancestry, he may pick "European" although his facial features are more typical of people from some other region of the globe.  There's also plenty of ethnic diversity in Europe (for example, millions of people of Asian ancestry in the UK). People are not terribly careful in filling out medical questionaires (particularly those with very many questions); self-reporting is not a reliable way to exclude people; and we can imagine many people in Europe reporting themselves to be of "European ancestry" regardless of where their parents came from.  Although the paper gives hints here and there trying to suggest that it used a sample of limited ethnic diversity, based on its content we should assume it analyzed a data set that included a rather full range of human ethnic diversity.  

Having identified such genetic variations (called SNP's) very weakly associated with gender and ethnic differences, the study called them "genomic regions associated with normal-range facial variation," even though there was no reason for thinking that such genomic regions were anything like a specification of a human face.  A similar technique would be to have a computer analyze address data and building descriptions.  You might find that variations in zip code were weakly associated with variations in building structures. For example, a zip code from the East Side of Manhattan (such as 10028) would be  associated with high building heights, and a zip code from rural Montana would be associated with low building heights. But zip codes do absolutely nothing to specify how tall a building should be or what type of structure it has, and it would be a most ridiculous falsehood to speak of a "zip code architecture for apartment buildings."  Similarly, genes do nothing to specify the structure of any visible part of a human face, so it is very misleading to be speaking of a "genetic architecture of the human face." It is a huge mistake to be confusing a mere association or correlation with a specification or an architecture, particularly when you merely have the type of extremely weak correlations reported by the study in question.  

The authors of the study have done nothing to support their use of the very erroneous phrase "genetic architecture of the human face."  We know why there can be no such thing as a genetic architecture of the human face, or any other piece of anatomy. Genes do not specify anatomy.  Genes do not specify body plans or the structure of skeletons, appendages or organs. Gene do not even specify the structure of cells. Genes merely specify low-level chemical information such as the sequence of amino acids that make up a polypeptide chain that is the starting point of a protein molecule. Were we to examine the genes mentioned in the paper, and look at the protein molecules that correspond to such genes, we would find structures that look nothing at all like a human face, and look nothing like any visible part of a human face. 

Here are a few relevant quotes by authorities, which collectively deny any claim that there is any such thing as a "genetic architecture of the human face":

• On page 26 of the recent book The Developing Genome, Professor David S. Moore states, "The common belief that there are things inside of us that constitute a set of instructions for building bodies and minds -- things that are analogous to 'blueprints' or 'recipes' -- is undoubtedly false."
• Biologist Rupert Sheldrake says this "DNA only codes for the materials from which the body is constructed: the enzymes, the structural proteins, and so forth," and "There is no evidence that it also codes for the plan, the form, the morphology of the body."
• Describing conclusions of biologist Brian Goodwin, the New York Times says, "While genes may help produce the proteins that make the skeleton or the glue, they do not determine the shape and form of an embryo or an organism." 
• Professor Massimo Pigliucci (mainstream author of numerous scientific papers on evolution) has stated  that "old-fashioned metaphors like genetic blueprint and genetic programme are not only woefully inadequate but positively misleading."
• Neuroscientist Romain Brette states, "The genome does not encode much except for amino acids."
• In a 2016 scientific paper, three scientists state the following: "It is now clear that the genome does not directly program the organism; the computer program metaphor has misled us...The genome does not function as a master plan or computer program for controlling the organism; the genome is the organism's servant, not its master.
• In the book Mind in Life by Evan Thompson (published by the Belknap Press of Harvard University Press) we read the following on page 180: "The plain truth is that DNA is not a program for building organisms, as several authors have shown in detail (Keller 2000, Lewontin 1993, Moss 2003)."
• Developmental biologist C/H. Waddington stated, "The DNA is not a program or sequentially accessed control over the behavior of the cell."
•  Scientists Walker and Davies state this in a scientific paper: "DNA is not a blueprint for an organism; no information is actively processed by DNA alone...DNA is a passive repository for transcription of stored data into RNA, some (but by no means all) of which goes on to be translated into proteins."
• Geneticist Adam Rutherford states that "DNA is not a blueprint," a statement also made by biochemistry professor Keith Fox.
• "The genome is not a blueprint," says Kevin Mitchell, a geneticist and neuroscientist at Trinity College Dublin, noting "it doesn't encode some specific outcome."
• "DNA cannot be seen as the 'blueprint' for life," says Antony Jose, associate professor of cell biology and molecular genetics at the University of Maryland, who says, "It is at best an overlapping and potentially scrambled list of ingredients that is used differently by different cells at different times."  
• Sergio Pistoi (a science writer with a PhD in molecular biology) tells us, "DNA is not a blueprint," and tells us, "We do not inherit specific instructions on how to build a cell or an organ." 
• Michael Levin (director of a large biology research lab) states that "genomes are not a blueprint for anatomy," and after referring to a "deep puzzle" of how biological forms arise, he gives this example: "Scientists really don’t know what determines the intricate shape and structure of the flatworm’s head."
• Ian Stevenson M.D. stated "Genes alone - which provide instructions for the production of amino acids and proteins -- cannot explain how the proteins produced by their instructions come to have the shape they develop and, ultimately, determine the form of the organisms where they are," and noted that "biologists who have drawn attention to this important gap in our knowledge of form have not been a grouping of mediocrities (Denton, 1986; Goldschmidt, 1952; B. C. Goodwin, 1985, 1988, 1989, 1994; Gottlieb, 1992; Grasse, 1973; E. S. Russell...Sheldrake, 1981; Tauber and Sarkar, 1992; Thompson, 1917/1942)."
• Biologist B.C. Goodwin stated this in 1989: "Since genes make molecules, genetics...does not tell us how the molecules are organized into the dynamic, organized process that is the living organism."
• An article in the journal Nature states this: "The manner in which bodies and tissues take form remains 'one of the most important, and still poorly understood, questions of our time', says developmental biologist Amy Shyer, who studies morphogenesis at the Rockefeller University in New York City."
• Timothy Saunders, a developmental biologist at the National University of Singapore. says, "Fundamentally, we have a poor understanding of how any internal organ forms.”
• On the web site of the well-known biologist Denis Noble, we read that "the whole idea that genes contain the recipe or the program of life is absurd, according to Noble," and that we should understand DNA "not so much as a recipe or a program, but rather as a database that is used by the tissues and organs in order to make the proteins which they need."
• A paper by Stuart A. Newman (a professor of cell biology and anatomy) discussing at length the work of scientists trying to evoke "self-organization" as an explanation for morphogenesis states that "public lectures by principals of the field contain confidently asserted, but similarly oversimplified or misleading treatments," and says that "these analogies...give the false impression that there has been more progress in understanding embryonic development than there truly has been." Referring to scientists moving from one bunk explanation of morphogenesis to another bunk explanation, the paper concludes by stating, "It would be unfortunate if we find ourselves having emerged from a period of misconceived genetic program metaphors only to land in a brave new world captivated by equally misguided ones about self-organization."
• Referring to claims there is a program for building organisms in DNA, biochemist F. M. Harold stated "reflection on the findings with morphologically aberrant mutants suggests that the metaphor of a genetic program is misleading." Referring to  self-organization (a vague phrase sometimes used to try to explain morphogenesis), he says, "self-organization remains nearly as mysterious as it was a century ago, a subject in search of a paradigm." 
• Writing in the leading journal Cell, biologists  Marc Kirschner, John Gerhart and Tim Mitchison stated, "The genotype, however deeply we analyze it, cannot be predictive of the actual phenotype, but can only provide knowledge of the universe of possible phenotypes." That's equivalent to saying that DNA does not specify visible biological structures, but merely limits what structures an organism can have (just as a building parts list merely limits what structures can be made from the set of parts). 
• A paper co-authored by a chemistry professor (Jesper Hoffmeyer) tells us this: "Ontogenetic 'information,' whether about the strucure of the organism or about its behavior, does not exist as such in the genes or in the environment, but is constructed in a given developmental context, as critically emphasized, for example, by Lewotin (1982) and Oyama (1985)."
• Biologist Steven Rose has stated, "DNA is not a blueprint, and the four dimensions of life (three of space, one of time) cannot be read off from its one-dimensional strand."
• At the Biology Stack Exchange expert answers site, someone posted a question asking which parts of a genome specify how to make a cell (he wanted to write a program that would sketch out a cell based on DNA inputs).  An unidentified expert stated that it is "not correct" that DNA is a blueprint that describes an organism, and that "DNA is not a blueprint because DNA does not have instructions for how to build a cell." No one contradicted this person's claim, even though the site allows any of its experts to reply. 
• "DNA is not a blueprint for an organism," states Templeton Prize-winning physicist and astrobiologist P. C. W. Davies.
• Two scientists said this: "We see no valid use for definitions of the genotype and phenotype in terms of blueprints, programs, or sets of instructions, and their realizations or manifestation....The program/manifestation metaphor is factually misleading, because it suggests that the genotype uniquely determines an organism’s phenotype. However, as is well known, all it does is specify an organism’s norm of reaction to environmental conditions (Rieger et al., 1991,Lewontin, 1992)."
• A 2022 paper in the journal Science (one authored by more than ten scientists) says this: "Although the genome is often called the blueprint of an organism, it is perhaps more accurate to describe it as a parts list composed of the various genes that may or may not be used in the different cell types of a multicellular organism....The genome in and of itself does not provide an understanding of the molecular complexity of the various cell types of that organism."

This is the right hand of biology, telling us loud and clear that genes do not specify human anatomy. But the left hand of biology keeps misleading us, by continuing to repeat or suggest the untrue claim that genes specify how to make a human body. 

Let us imagine a small child trying to explain how astronauts got to the moon. He might tell us some silly story like this:

"I heard guys walked on the moon. I know how they must have done it. They must have got a real big balloon, kind of like the ones at birthday parties, but much bigger. Then they must have rode the balloon all the way to the moon."

Helium balloons rise up into the air because helium is lighter than the air, but a helium balloon would be of no value in crossing the airless space between Earth and the moon.  And no one ever tried to ride a balloon to the moon. So the story told by the child would be both false and childish.  Similarly, the story told by many a biologist (that the incredibly organized arrangements of matter in organisms arise from blueprints in DNA) is a story that is both false and very childish. The story is false because there is no blueprint of human anatomy in genes or DNA. The story is very childish because blueprints don't build things. Things can only get built from blueprints when there are intelligent agents such as construction workers who read the blueprints, understand the blueprints, and use the blueprints to construct complex things. Not only is there nothing in the human body like an anatomy blueprint, there is also nothing below the neck that could act as a blueprint reader and "make what the blueprint specifies" if a blueprint for human anatomy happened to exist inside DNA. 

A respectable way to do a study such as the "Insights into the Genetic Architecture of the Human Face" would be:

(1) Before doing any work, publish a very exact research plan discussing precisely how data would be gathered and analyzed, and then follow such a plan to the letter.

(2) Create a strict blinding protocol to be followed to reduce the chance of experimental bias.

But the study mentions no blinding protocol, and makes no mention of any pre-registration of a research plan. Instead the paper rather suggests  the authors were improvising as they went along in order to get the desired result, playing around with various exotic statistical contortions until they got the desired result, and slicing and dicing the data until it produced the desired result. For example, we read the following (I have put some words in boldface):

"First, for each segment, we separated the 203 variants into three groups and ran three SEM models on each of these groups, plus all covariates. If any of the three subset SEMs did not converge, we then re-grouped the SNPs into four or more groupings and re-ran the subset SEM models on these groupings. This process was repeated until all subset SEMs converged and we had parameter estimates for all 203 SNPs. Next, for each segment, SNPs with p-values lower than 0.2 in the initial subset SEMs were collected and a unified SEM model for each segment was created and subsequently refined. If the unified SEM model did not converge, then this segment was discarded and no further analysis was performed. If all of the SNPs included in the unified model had p-values lower than 0.2, a cutoff selected to maintain model stability, no further changes were made, and we reported the model fit indices and parameter estimates. For segments where the unified SEM model produced SNP p-values greater than 0.2, the SNPs included in the SEM model were pruned by selecting SNPs with p < 0.05 and the model was re-run with this reduced set of SNPs. This process was repeated until all SNPs had p-values lower than 0.2. In the case of segments 7, 16, and 25, this iterative pruning process led to a rapidly declining model, so we elevated the SNP pruning p-value from 0.05 to 0.1 to account for instability in these models....In general, the number of model parameters generated by the final refined SEM model for each segment ranged between 92 and 217, depending on the number of shape PCs and SNPs included in each model. As 8,246 participants were used, this led to a range of 38-90 participants per parameter, which is well above recommendations. Additional statistical power was lent to our models by having a large number of samples and input variables per latent factor. Of the 63 segments, the SEM models for 13 segments were discarded because they did not converge on a solution, which normally occurs when variants are non-informative for that particular segment or the variance of the segment is low. For each of the 50 SEM models where the refinement process was successful, we evaluated the fit of each model by instituting cutoffs on the following indices: Chi-square (p-value < 0.05), comparative fit index (CFI > 0.90), root mean square error of approximation (RMSEA < 0.08), and standardized root mean square residual (SRMR < 0.08), which generally indicate the strength of how well the SEM models the data."

As the old saying goes, "Torture the data sufficiently, and it will confess to anything."

Occasional very weak statistical associations can always be found between totally unrelated things, by a person willing to look in enough places, and willing to engage in vigorous statistical manipulations until he finds what he is looking for. A sufficiently energetic and motivated analyst could probably analyze the genome of the sunflower plant (which is actually larger than the human genome), and find a few hundred genes that had a weak statistical association between features of the human face.  Given two sets of unrelated data, and a willingness to play around with statistical tricks until you find a few weak correlations you are hoping to find, analysts can always find some weak correlations here and there between totally unrelated data sets, simply because of a few random mathematical coincidences found when comparing two large unrelated data sets. 

The authors of the "Insights into the Genetic Architecture of the Human Face" claim to have found some very weak correlations between 1% of the human genome and some facial database.  Using the same "pull out all the stops"  and "keep trying until something coughs up" statistical techniques, you could find weak correlations between 1% of the daily numbers in things like this:

• 1960's price movements of Brazilian soybeans and 1990's yields of New England cod fishermen;
• 1960's stock price movements of European electronics manufacturers and 1990's rainfall fluctuations in Africa;
• 1950's American high school students grades and 1990's European soccer scores.

Postscript: The term "body plan" is a profoundly misleading term that biologists love to use, a term that opens the door to deceptions about DNA. In biology literature the term "body plan" has a very limited meaning, something vastly different from a complete plan for constructing an organism. According to a scientific paper "a body plan is a suite of characters shared by a group of phylogenetically related animals at some point during their development." The wikipedia.org article on "body plan" tells us this: "A body plan, Bauplan (German plural Baupläne), or ground plan is a set of morphological features common to many members of a phylum of animals." 

According to this definition, all chordates (including men, bears, dogs and fish) have the same body plan. So when biologists talk about "the human body plan" they

are merely referring to the common characteristics of all chordates, including men, bears, dogs and fish:  basically just the existence of a backbone and bilateral symmetry (having the same things on both sides of the body).  They are not referring to the structure of the 200 types of cells in the human body, or the structure of internal organs, and are not referring to the dynamic intricacies of human physiology. But anyone hearing the term "body plan" will think the term referred to a complete specification of a human body.  So, most misleadingly, biologists may say that this or that "determines the body plan," when all they mean is the beginning of a bilateral organism with a backbone, something a thousand times simpler than the final product of the internally dynamic and enormously organized human body.  This is as misleading as someone saying that he has built a starship, when he has merely built a boat in the shape of a star.