Discussion of ЭТЭ with Anatolii Protopopov

Anatoliy Protopopov has posted on his website a commentary on my column about ETE. I am copying this commentary here so that I can respond to it. An informational perspective on the epigenetic theory of evolution. Open letter to Dmytro Shabanov Dmytro Andriyovych! Your recent columns in "Kompyuterra" regarding ETE have given me serious cause for reflection. To such an extent that my considerations, initially begun as a response on the forum, have grown into this rather lengthy text, hardly suitable for forum discussion. I have read the authors you recommended. Although I will not claim to have carefully read every word, I believe I have understood the general principles. Nevertheless, first and foremost, it seems necessary to set out my understanding of the principles of CET and ETE: perhaps I am simply misunderstanding something. So, according to CET, the phenotype is not inherited. Starting from the genotype, the phenotype can vary substantially under the influence of the specific conditions of ontogenesis at the time, but these variations will not be transmitted to the next generation. It is acknowledged that the peculiarities of the cytoplasm of the maternal egg cell, including those acquired during the mother's ontogenesis, can somehow influence the ontogenesis of offspring, but this influence is considered insignificant, non-specific, and not extending beyond one generation. In any case—I am not aware of any cases of firmly established influence of specific properties of egg cell cytoplasm (chemistry, structure) on phenotypic specifics. Only variations (mutations, etc.) of the genotype of the germ line cells of the organism are transmitted to the next generation. According to ETE, certain phenotypic peculiarities (namely—the degree and direction of its deviation from the "reaction norm" determined by the genotype) can be transmitted to the next generation. That is, actually inherited; moreover, initially by some "epigenetic" (epigenetic, that is) mechanisms, not changing the genetic code proper, and in some such special manner, allowing one not to disturb the spirit of Jean-Baptiste Lamarck in vain. Further, if the deviation from the "genotypic norm" is repeated in subsequent generations, it will become fixed in the genotype. In effect, the theory postulates the existence of some "operational" heredity, not immediately affecting the main genetic code ("strategic" heredity), but which can, nevertheless, be transmitted to at least one or two subsequent generations. The problem lies in the specific mechanisms of such epigenetic inheritance. You have stated that such details are not important for ETE; I find this surprising. After all, if one does not propose some, at least somewhat convincing, mechanism, the entire theory hangs in the air. What might these mechanisms be? The simplest is in unicellular organisms reproducing by simple division: there the phenotype is inherited "by itself"—and even that is not entirely the case. But if one proceeds through the zygote, the properties of the offspring can be inherited only by two means—through the properties of the egg cell (and almost exclusively—its cytoplasm), or through genetic material. However, the cytoplasm of the zygote is exclusively maternal! This alone gives cause for reflection—for in ETE there is no (at least—I have not noticed) emphasis on epigenetic inheritance along the maternal line. Furthermore, if for genes it is more or less (at least in principle) clear how a change in DNA structure might be reflected in the development of certain traits in offspring, then similar data on cytoplasm are very sparse and unconvincing. Regarding genetic epigenetic inheritance, there is slightly more clarity. For example, there is so-called "DNA methylation." It indeed does not affect the main genetic code, yet it influences gene activity, and accordingly—the phenotype. And can, it seems, be produced "in the course of work," which theoretically opens the possibility of operatively responding to the current situation. What personally concerns me is the fact that in the mammalian genome, the greater part of CG nucleotide pairs is methylated, which greatly reduces the possibility of transmitting unique information through methylation, as well as the unclear mechanism of recoding operatively recorded "notes" onto the "tablets" of the main code—but never mind. Consider the question more broadly—from the standpoint of informatics in general. It is known that the genome of any multicellular being, in the informational sense, is surprisingly compact. 10 Mb (and even 700) cannot possibly contain sufficient information about building an organism "from scratch." However, no organism is built "from scratch"—it is built on the basis of an already existing and functioning cell. This is the greatest blessing—for extremely important and extensive information about "how to build a cell from scratch from available materials and breathe life into it" can be transmitted hereditarily, limited to some working blueprints. Rather—transmit, but not information, so to speak, the nature. But the fact does not save the situation much given the genome's sparseness. Organisms are just too complex. Dawkins (and not only he) resolves this by warning against likening the genome to a blueprint, and recommends likening it to a culinary recipe. This is true, yes, but there are no miracles. However one spins it, the genome must contain sufficient information about building an organism, whatever one calls it—a blueprint, a recipe, or otherwise. What, in the informational sense, distinguishes a blueprint from a recipe? Obviously, compactness. A blueprint immediately shows the future product; from a recipe this, generally speaking, is not evident. It must first be "executed," i.e., the prescribed sequences of actions must be performed, and only then can one see what resulted. Exactly the same happens in computer archiving-dearchiving programs. The archive content must be fully processed, and only then will the required information be accessible. Moreover—even the scheme of the computer archiver's operation has much in common with executing a culinary recipe, and further—with the development of an organism from a zygote! A recipe may state: take so much flour, butter, sugar, salt, and do this and that to them. Recipes generally do not explain what flour is, what sugar is, or how to obtain them from source materials—and this is a very great economy in information volume! The same we have with the zygote—the source cell already exists, and in the "explanations" it does not need to be described. But a cookbook may also contain recipes for creams, marinades, and sauces, which in explanations still need explanation, and these explanations are there. However, as a rule, only once. Further, various recipes may contain only a reference to this or that sauce or marinade. The latter approach is the main one in the work of classical computer archivers. As the source file is processed, a so-called "dictionary" is compiled, containing the most frequently occurring byte combinations, and the archive already contains only the position numbers in this dictionary. Clearly, looking at the archive content, we can say absolutely nothing about the original information: only complete unpacking, and then... If the archiver is single-pass (and such are the absolute majority), then the composition of this dictionary will be uneven for different parts of the file: some dictionary entries may become unnecessary and undergo "apoptosis," some will be added, to perhaps disappear later as well. The similarity with embryonic development is obvious: in the course of embryogenesis, many temporary structures and environments are also generated, which are used as source material for further steps, which may undergo apoptosis once the need for them has passed. All of this—compactness, multi-stage nature, and the rest—allows us to speak of the "archived" nature of multicellular genomes. In what other ways is a computer archive similar to a genome? In the specifics of making changes to the target information. One cannot make such changes directly in an archive that would meaningfully and purposefully change the target information. This applies not only to universal archivers, but also to all other methods of information compression—for example, in JPEG images. To edit a JPEG image, it must be fully unpacked, converted to "blueprint" form (so-called Bitmap), changes made to this "blueprint," and then repacked into JPEG. Graphic editors do this automatically, which is why not all users are aware of it; nevertheless, it is so, and precisely because of this, when working with an image in multiple stages, it is not recommended to save intermediate stages in JPEG, since JPEG is "lossy" compression, and losses accumulate at each stage. But for the genome, there is no "packing" procedure! Gamete precursor cells separate at a fairly early stage of embryogenesis, and their genome contains nothing differing in any way from the genome of the zygote. No mechanism is known that would "scan" the current state of the organism and "pack" it into the genetic code—whether methylated or not—into germ line cells. Oh yes, methylation. It indeed can occur later, but in what exact place? Well, of course, if one thinks very, very hard, runs a simulation model on a computer, performs other highly complex calculations, then perhaps one can indicate the place. But are you sure this is within the capabilities of an organism, say, a rotifer? The most valuable thing for us in this analogy is that we can very easily conduct the corresponding experiments. Not on the genome (that is expensive and very time-consuming), but on a computer archive. For this, all that is required is a file editor that can edit bytes in a file directly, without trying to interpret them in any way. I, for example, used the editor built into the FAR manager, but any other can be used. Unfortunately, universal archivers, like RAR, very strictly control the integrity of their archives, so any attempt to directly distort some bytes will cause a checksum failure, and the archiver will refuse to work with that archive. More democratic in this respect are JPEG images; such "mutations" can already be introduced into them. And although their principle of operation differs more greatly from the genetic one, attempts to introduce "mutations" cause a much more "biological" effect. Namely—most "mutations" will be neutral, or nearly neutral, however, occasionally catastrophic distortions of the image will be observed. Moreover, even if the changes are slight (but evident), it will be impossible to predict in advance what exactly and how will change. The purity of the experiment is somewhat marred by the fact that a JPEG image consists of a regular array of 8x8 pixel squares (and predicting the location to within this area is not difficult), but there is its own aesthetic charm in this. Try it—it is not difficult. And instructive! If, that is, one agrees that the parallels with the genome are correct.