Can natural selection create?

Can natural selection create?

Newsflash - Mutation/selection cannot even create a single gene.

by Dr.J.C.Sanford

We have been analyzing the problem of genomic degeneration and we have found that regardless of how we analyze it, the genome must clearly degenerate. This problem overrides all hope for the forward evolution of the whole genome. However some limited traits might still be improved via mutation/selection. Just how limited is such progressive ("creative") mutation/selection? From the perspective of our analogy, an instruction manual, we can intuitively see that not even a single component of a jet plane (let's say a molded aluminum component) could realistically arise by misspellings within the manual. So it is certainly reasonable to then ask the parallel question - "Could mutation/selection create even a single functional gene?" The answer is that it cannot - because of the enormous preponderance of deleterious mutations, even within the context of a single gene. The net information must always still be declining, even within a single gene. However, to better understand the limits of forward selection, let us for the moment discount all deleterious mutations and only consider beneficial mutations. Could mutation/selection then create a new and functional gene?

1. Defining our first desirable mutation - The first problem we encounter in trying to create a new gene via mutation/selection is defining our first beneficial mutation. By itself, no particular nucleotide (A,T,C, or G) has more value than any other - just as no letter in the alphabet has any particular meaning outside of the context of other letters. So selection for any single nucleotide can never occur, except in the context of all the surrounding nucleotides (and in fact within the context of the whole genome). Like changing a letter within a word or chapter, the change can only be evaluated in the context of all the surrounding letters. We cannot define any nucleotide as good or bad except in relation to its neighbors and their shared functionality. This brings us to an excellent example of the principle of "irreducible complexity". In fact, it is irreducible complexity at its most fundamental level. We immediately find we have a paradox. To create a new function, we will need to select for our first beneficial mutation, but we can only define that new nucleotide's value in relation to its neighbors. Yet to create any new function, we are going to have to be changing most of those neighbors also! We create a circular path for ourselves - we will keep destroying the "context" we are trying to build upon. This problem of the fundamental inter-relationship of nucleotides is called epistasis. True epistasis is essentially infinitely complex, and virtually impossible to analyze, which is why geneticists have always conveniently ignored it. Such bewildering complexity is exactly why language (including genetic language) can never be the product of chance, but requires intelligent design. The genome is literally a book, written literally in a language, and short sequences are literally sentences. Having random letters fall into place to make a single meaningful sentence, by accident, is numerically not feasible. The same is true for any functional strings of nucleotides. If there are more than several dozen nucleotides in a functional sequence, we know that realistically they will never just "fall into place". This has been mathematically demonstrated repeatedly. But as we will soon see, neither can such a sequence arise randomly one nucleotide at a time. A pre-existing "concept" is required as a framework upon which a sentence or a functional sequence must be built. Such a concept can only pre-exist within the "mind of the author". Starting from the very first mutation, we have a fundamental problem - even in trying to define what our first desired beneficial mutation should be!

2. Waiting for the first mutation - Human evolution is generally assumed to have occurred in a small population of about 10,000. The mutation rate for any given nucleotide, per person per generation is exceedingly small (only about one chance in 30 million). So in a typical evolutionary population, if we assume 100 mutations per person per generation, one would have to wait 3,000 generations (at least 60,000 years) to expect a specific nucleotide to mutate - within a population of 10,000. But two out of three times, it will mutate into the "wrong" nucleotide. So to get a specific desired mutation at a specific site will take three times as long - or at least 120,000 years. Once the mutation has occurred, it has to become fixed (such that all individuals in the population will have two copies of it). For new mutations, because they are so rare within the population, they have an extremely great probability of being lost from the population, due to random genetic drift. Only if the mutation is dominant and has a very distinct benefit does selection have any reasonable chance to rescue any given new mutation from random elimination via drift. According to population geneticists, apart from effective selection, in a population of 10,000 our given new mutant has only one chance in 20,000 (the total number of non-mutant nucleotides present in the population) of not being lost via drift. Even with some modest level of selection operating, there is a very high probability of random loss, especially if the mutant is recessive or is weakly expressed (we actually know that almost all beneficial mutations will be both recessive and nearlyneutral). For example, if a mutation increases fitness by half of one percent, it only has a 1% probability of becoming fixed. So realistically, at least 99 out of 100 times the desired beneficial mutation will be randomly lost. So a typical mildly-beneficial mutation must happen about 100 times before it is likely to "catch hold" within the population (even though it is beneficial!). So on average, we would have to wait 120,000 x 100 = 12 million years to stabilize our typical first desired beneficial mutation, to begin building our hypothetical new gene. So, in the time since we supposedly evolved from chimp-like creatures (6 million years), there would not be enough time to realistically expect our first desired mutation - the one destined for fixation.

3. Waiting for the other mutations - After our first mutation has been found (the one that will eventually be fixed), we need to repeat this process for all the other nucleotides encoding our hoped-for gene. A gene is minimally 1,000 nucleotides long (this is really 50-fold too generous - I am ignoring all regulatory elements and introns). So if this process was a straight, linear, and sequential process - it would take about 12 million years x 1,000 = 12 billion years to create the smallest possible gene. This is approximately the time since the reputed big bang! So it is a gross understatement to say that the rarity of desired mutations limits the rate of evolution!

4. Waiting for recombination - Because sexual species (such as man) can shuffle mutations, it might be thought that all the needed mutations for a new gene might be able to occur simultaneously within different individuals within the population, and then all the desirable mutations could be "spliced together" via recombination. This would mean that the mutations would not have to occur sequentially - shortening the time to create the hoped-for gene (so we might need less than billions of years). There are two problems with this. Firstly, when we examine the human genome, we consistently find the genome exists in large blocks (20,000-40,000 nucleotides) wherein no recombination has occurred - since the origin of man (Gabriel et al. 2002, Tishkoff and Verrelli, 2003). This means that virtually no meaningful shuffling is occurring on the level of individual nucleotides. Only large gene-sized blocks of DNA are being shuffled. I repeat - no actual nucleotide shuffling is happening! Secondly, even if there were effective nucleotide shuffling, the probability of getting all the mutants within the population to shuffle together into our hoped-for sequence of 1,000 is so astronomically remote that we would need even more time than by the sequential approach (even more billions of years) for this scenario to work. Lastly, if there really were this type of extensive "nucleotide shuffling", which might build a new gene in this way, the very first generation after the new gene fell into place, it would be torn apart again by the same extensive nucleotide shuffling. In poker, it is not likely you will be dealt a royal flush. If you are, and then the cards are reshuffled - what are the odds you will then get that very same hand dealt to you again?

5. Waiting on "Haldane's dilemma" - Once that first mutation that is destined to become fixed within the population has finally occurred, it needs time to undergo selective amplification. A brand new mutation within a population of 10,000 people, exists as only one nucleotide out of 20,000 alternatives (there are 20,000 nucleotides at that site, within the whole population). The mutant nucleotide must "grow" gradually within the population, either due to drift or due to natural selection. Soon there might be two copies of the mutant, then 4, then 100, and eventually - 20,000. How long does this process take? For dominant mutations, assuming very strong unidirectional selection, the mutant might conceivably "grow' within the population at a rate of 10% per generation. At this very high rate, it would still take roughly 105 generations (2,100 years) to increase from 1 to 20,000 copies (1.1^105 = 20,000). However, in reality mutation fixation takes very much longer than this, because selection is generally very weak, and most mutations are recessive and very subtle. When the mutation is recessive, or when selection is not consistently unidirectional or strong, this calculation is much more complex - but it is obvious that the fixation process would be very dramatically slower. For example, an entirely recessive beneficial mutation, even if it could increase fitness by as much as 1%, would require at least 100,000 generations to fix (Patterson, 1999).

A famous geneticist, Haldane (1957), calculated that given what he considered a "reasonable" mixture of recessive and dominant mutations, it would take (on average) 300 generations (at least 6,000 years) to select a single new mutation to fixation. Selection at this rate is so very slow, it is essentially the same as no selection at all. This problem has classically been called "Haldane's dilemma". At this rate of selection, one could only fix 1,000 beneficial nucleotide mutations within the whole genome, in the time since we supposedly evolved from chimps (6 million years). This simple fact has been confirmed independently by Crow and Kimura (1970), and ReMine (1993, 2005). The nature of selection is such that selecting for one nucleotide always reduces our ability to select for other nucleotides (selection interference) - therefore simultaneous selection does not hasten this process.

At first glance, the above calculation seems to suggest that one might at least be able to select for the creation of one small gene (of up to 1,000 nucleotides) in the time since we reputedly diverged from chimpanzee. There are two reasons why this is not true. First, Haldane's calculations were only for independent, unlinked mutations. Selection for 1,000 specific and adjacent mutations could not happen in 6 million years - because that specific sequence of adjacent mutations would never arise - not even in 6 billion years. One cannot select mutations that have not happened. Secondly, as we will soon see, the vast bulk of a gene's nucleotides are near-neutral and cannot be selected at all - not in any length of time. The bottom line of Haldane's dilemma is that selection to fix new beneficial mutations occurs at glacial speeds, and the more nucleotides which are under selection, the slower the progress. This severely limits progressive selection. Within reasonable evolutionary timeframes, we can only select for an extremely limited number of unlinked nucleotides. In the last 6 million years, selection could maximally fix 1,000 unlinked beneficial mutations - creating less new information than is on this page of text.* (* In terms of information content, 3 nucleotides equal roughly 1 typewritten character (there are only 4 nucleotides, but 26 letters, and more than 64 keys on a keyboard). So one codon triplet equals roughly one typographical "letter", and thus 1000 nucleotides equals only 333 spaces on a typewritten page.) There is no way that such a small amount of information could transform an ape into a human.

Although we have temporarily suspended deleterious mutations from consideration, it is only fair now to note that within the same timeframe that we hypothetically evolved from chimps, geneticists believe that many thousands of deleterious mutations should have been also fixed, via genetic drift (Kondrashov, 1995; Crow, 1997; Eyre-Walker and Keightley, 1999; Higgins and Lynch, 2001). Therefore, our evolutionary assumptions should lead us to logically conclude that we should have significantly degenerated downward from our ape-like ancestors (deleterious fixations greatly outnumbering beneficial fixations). The power of this logic is overwhelming. In fact, we know man and chimp differ at roughly 150 million nucleotide positions (Britten, 2002), which are attributed to at least 40 million hypothetical mutations. Therefore, assuming man evolved from a chimp-like creature - during that process there must have been about 20 million nucleotide fixations within the human lineage (40 million divided by 2), yet we now can see that natural selection could only have selected for 1,000 of these mutations. All the rest (about 20 million) would have had to have been fixed by random drift - resulting in millions of nearly-neutral deleterious substitutions. The result? A maximum of 1000 beneficial substitutions - in opposition to millions of deleterious substitutions. This would not just make us inferior to our chimplike ancestors, it would obviously have killed us!

Next pag