Dillon & Cooper on deleterious mutation rates

I just read Dillon & Cooper 2016, an article that was extensively and “robustly” discussed in a recently closed thread. I believe that 2 key points of the article were overlooked. I would like to raise them here for the benefit of the community.

The first point is that this mutation accumulation (MA) experiment used a methodology that excluded the effects of selection:

Nearly all prior research on the fitness effects of mutations has studied mutants that have been screened by selection, which presumably purged deleterious variants and enriched beneficial ones. MA experiments are designed to minimize effects of selection to the greatest extent possible, thus capturing most mutations independent of the biases of natural selection.

The key point here is that the result of the experiment does not, in the least, represent the effects of mutation in the real world.

Secondly, Sanford’s Genetic Entropy (GE) hypothesis as I understand it would expect two outcomes:

  1. None of the 43 lineages would be more fit than the original population, and
  2. Many would be less fit.

The problem with @PDPrice’s analysis is that he only pays attention to the second of the two predictions. And indeed, in the absence of selection, the mutations were biased toward reduced fitness at the end of the experiment. Paul then cites the study’s conclusion about preponderance of deleterious mutations as strong evidence for genetic entropy.

The problem with this analysis is that it ignores the fact that no fewer than 4 of the experiment’s lineages had gained a significant fitness advantage over the ancestral population.

in M9MM … we observed 13 lineages with significantly reduced fitness in M9MM, but also four lineages with significantly increased fitness [my emphasis]

This is the evidence that strongly refutes the GE hypothesis. In a real world environment, the great number of less fit strains would be selected against, certainly. Simultaneously, the fitter strains would be positively selected. Thus the fitter strains (as defined by the fitness gradients in their environments) would predominate.

And how do we know this? You have to exclude natural selection to even conduct an experiment like Dillon and Cooper’s. Once you introduce natural selection into an experiment, the ratio of positive to negative mutations preserved at the end is far more beneficial for the surviving population.

I invite biologists such as @T_aquaticus, @Mercer, @glipsnort, and @swamidass to clarify or correct any misunderstandings.

Chris Falter


The problem with your objection is that @PDPrice has abandoned fitness as a criterion and is now embracing “information loss” and “devolution”, which have the advantage of being nebulous, impossible to quantify, and uninvestigated.

All of that has been discussed before. Given the volume of the threads, you can be forgiven for overlooking that discussion.


It is worth noting how natural selection is reduced in this experiment (if I understand it correctly). What they did was take 75 bacteria from a single culture and use those single bacteria to found each lineage. Each day, they took one bacteria from that population to propogate the lineage. This means there is an extreme bottleneck every 25 generations or so. This will fix deleterious mutations since there isn’t enough time or individuals to remove deleterious mutations. If a single cell carries a deleterious mutation and is the single founder of the next generation then that deleterious mutations will be found in all bacteria over the next 25 generations, and much longer.

Also worth noting is bacterial genomes are very gene dense, especially compared to eukaryotic genomes:

By my count, That’s about 7,000 genes found in 7.7 million bases, or one gene per 1,000 bases. Compare this to humans who have about 30,000 genes across 3 billion bases, or one gene per 100,000 bases. That’s a 100 fold difference, and gene size (in exons) isn’t going to differ enough to really matter. For this reason, we would expect deleterious mutations to be much higher compared to neutral mutations in a gene dense genome, especially when comparing it to a genome where genes are more spread out.

Finally, if these bacteria were allowed to compete with one another instead of passing through extreme single cell bottlenecks then these deleterious mutations would be filtered out and the beneficial mutations would be amplified, as noted in the OP.


You are correct. To biologists, the misrepresentations by GE proponents are blatant in this and other cases.

You are correct. That was why I opened my questioning of @PDPrice on the complete misrepresentations of the influenza A H1N1 data the published graph in the web articles by pointing out that even without any misrepresentations, every influenza sequence in that figure came from a plaque-purified virus, just as the sequences in the paper you bring up come from isolates cloned in the lab.

Both isolations constitute selection in and of themselves.


No. The fact that an increase in fitness is observed in some lineages doesn’t refute the GE hypothesis. Not at all. I have made this point in another thread.

If I understand correctly, in GE theory positive fitness can be attained for a small time by removing genes or base pairs or something. So if the number of genes, bps etc. remains the same or increases, then this breaks the GE prediction.

Then what would?


A demonstration of a net gain of information

Define information in a way where we can quantify and measure it, and then show by example a loss, so that we can infer what you mean by a gain.


How would one demonstrate that? How is information measured?

1 Like

What about empirically observed cases of gene duplication followed by subsequent mutations to the copy which produce new function while still retaining old function? Why doesn’t that count as new information?

Gene duplication and the evolution of moonlighting proteins

Gene duplication is a recurring phenomenon in genome evolution and a major driving force in the gain of biological functions. Here, we examine the role of gene duplication in the origin and maintenance of moonlighting proteins, with special focus on functional redundancy and innovation, molecular tradeoffs, and genetic robustness. An overview of specific examples-mainly from yeast-suggests a widespread conservation of moonlighting behavior in duplicate genes after long evolutionary times. Dosage amplification and incomplete subfunctionalization appear to be prevalent in the maintenance of multifunctionality. We discuss the role of gene-expression divergence and paralog responsiveness in moonlighting proteins with overlapping biochemical properties. Future studies analyzing multifunctional genes in a more systematic and comprehensive manner will not only enable a better understanding of how this emerging class of protein behavior originates and is maintained, but also provide new insights on the mechanisms of evolution by gene duplication.



So the reverse of that is an increase in information, right? So basically any gene gain, even duplication, is the opposite of genetic entropy.


Is this information?


You didn’t make that point. You offered up a silly analogy with no connection to any actual biological process.

Do you agree the fact life has been on the planet for 3.5 billion years without going extinct is conclusive disproof of GE’s claims?


Yes, in the long run, a net increase in information would be the opposite of GE.

1 Like

How long is “the long run”?


Then you admit GE has already been falsified.

Gene evolution and gene expression after whole genome duplication in fish: the PhyloFish database


With more than 30,000 species, ray-finned fish represent approximately half of vertebrates. The evolution of ray-finned fish was impacted by several whole genome duplication (WGD) events including a teleost-specific WGD event (TGD) that occurred at the root of the teleost lineage about 350 million years ago (Mya) and more recent WGD events in salmonids, carps, suckers and others. In plants and animals, WGD events are associated with adaptive radiations and evolutionary innovations. WGD-spurred innovation may be especially relevant in the case of teleost fish, which colonized a wide diversity of habitats on earth, including many extreme environments. Fish biodiversity, the use of fish models for human medicine and ecological studies, and the importance of fish in human nutrition, fuel an important need for the characterization of gene expression repertoires and corresponding evolutionary histories of ray-finned fish genes. To this aim, we performed transcriptome analyses and developed the PhyloFish database to provide (i) de novo assembled gene repertoires in 23 different ray-finned fish species including two holosteans (i.e. a group that diverged from teleosts before TGD) and 21 teleosts (including six salmonids), and (ii) gene expression levels in ten different tissues and organs (and embryos for many) in the same species. This resource was generated using a common deep RNA sequencing protocol to obtain the most exhaustive gene repertoire possible in each species that allows between-species comparisons to study the evolution of gene expression in different lineages. The PhyloFish database described here can be accessed and searched using RNAbrowse, a simple and efficient solution to give access to RNA-seq de novo assembled transcripts


Still waiting on a coherent, quantifiable definition of information. And what constitutes “the long run”?

Also, why must some particular gain of information stick around for “the long run”? The reality of biological evolution is not contingent on the claim that information gains(however you define and quantify that) will last indefinitely.