I subscribe to Nature Magazine, the "international weekly journal of science" which is basically weekly porn for scientists. This week's title is "Cache transactions" and under the very sexy picture of a HOTHEAD mutated plant.

The subtitle is "non-mendelian inheritance in arabidopsis" and it is apparently an important proofpoint in the world of evolution. My rough understanding of the paper significance is as such: basically DNA has inheritance mechanisms built in that goes beyond what you inherit from your dad and mom hence the "non-mendelian".

Here is the abstract:

Genome-wide non-mendelian inheritance of extra-genomic information in Arabidopsis. A fundamental tenet of classical mendelian genetics is that allelic information is stably inherited from one generation to the next, resulting in predictable segregation pattens of differing alleles. Although several exceptions to this principle are known, all represent specialized cases that are mechanistically restricted to either a limited set of specific genes (for example mating type conversion in yeast) or specific types of alleles (for example alleles containing tranposons or repeated seqauences). Here we show that Arabidopsis plants homozygous for recessive mutant alleles of the organ fusion gene HOTHEAD (HTH) can inherit allele-specific DNA sequence information that was not present in the chromosomal genome of their parents but was present in previous generations. This previously undescribed process is shown to occur at all DNA sequence polymorphisms examined and therefore seems to be a general mechanism for extra-genomic inheritance of DNA sequence information. We postulate that these genetic restoration events are the result of a template-directed process that makes use of an ancestral RNA-sequence cache.

In plain english, they detected alleles that can skip a generation. The alleles were not present in the N-1 generation, the physical allele, we are not just talking about a recessive trait here which would yield a 25% mutant population from a 0% expressed parent population, just like 2 brown eyed parent with grand parent with blue eyes have one chance in four to yield blue eyed babies. Here the physical bytes are gone in the N-1, were present in the N-2 and are present in the N generation. The population of mutant arabidopsis plants yielded a 3% reversion rate of the chromosome bearing HOTHEAD. That reversion is the result.

Then they *postulate* (no direct evidence of that mechanism) that essentially nature does "versioning" of its DNA. So that a random mutation can be reversed if it has negative effects. In the article they argue that metabolic stress is the feedback mechanism (the plant has all its organs fused together) and the plant knows "something is wrong" and will revert the DNA blueprint to the cached one. What is puzzling here for the genomists is that there is no direct evidence of the existance of this "master versioned copy" and it is not chromosomal DNA, by definition. Hence the postulate that the most likely candidate is some cache of RNA from the ancestors.

Basically they are zeroing in on the equivalent of a database for DNA :)

The reason I am so interested in genomics as they relate to evolution is that software and biology share the same interest in maintenance of complex systems. It used to be that biologists would come to software. I am tempted to do the reverse :) We know, in the software field, that the best way to maintain and evolve large systems is to modularize them by "aspectization". Yes here I come again with AO.

The first time I realized that all current mendelian/darwinistic approaches start *WITH A GRAPH OF SPECIES* that is very akin to the inheritance graph of object I immediately thought of aspects. We know that parent-child inheritance graphs are inherintly brittle when it comes to large OO systems maintenance. Nature is the perfect example of a complex system that needs continuous maintenance. For example how do you code for a new adaptive feature under stress. Wouldn't it be nice if this snippet of DNA was weaved in the mass by other means than inheritance, which is very slow. Mendelian inheritance in this scenario would basically KILL a large part of the population, that don't have the solution, except those that are lucky enough, through random mutation, to find it. Aspects deal with that distribution in software by being able to dynamically weave code in the original class/specie.

For example, let's take the case of a new SQL injection attack, the AO line of defense would be to deploy aspects to every JDBC drivers in your domain to patch the connections. The "parent" JDBC driver, may not have that feature built in its DNA but you can easily deploy the "good virus" to inject its protection pattern in the driver, simply filtering the SQL commands flowing through. This is by the way research we are following in academia.

One missing link becomes the packaging, we know how to deploy aspects in jars today in software, but exactly how does nature do that? The notion of "transposons" is an interesting one. It has been shown that DNA can self inject itself into a host. Think of it as the weavers (bytecode manipulation) of nature. Some research points to "viruses". Forget for a second the ones that get you sick or the ones that infect your laptop. 99.999% of viruses are actually benign. It has been proven that DNA from the virus could be weaved in the host DNA. That type of insertion of code is the same as what aspects do. In other words the advice code would fuse with the target code, essentially inlining, so that the subsequent copies (class) would have the data through inheritance even though it was weaved in extra-inheritance at some point.

I don't know, it makes a wholota sense to me :)

These theories are actually very controversial, kind of like AO, and not everyone is sold on them. But that is why I like science, what the paper proves for real is that there IS code weaving other than mendelian whatever that mechanism is. That today is a fact, we just don't know what it is yet. So have I been fascinated by this article and will be scouting for an interesting article along these lines in the future.

Anyway, I thought this different definition of "cache transactions" would amuse many of you, funny what nature has in store... Anyone want to put some money on the fact that she uses AO? :)

remember she loves you, or maybe she doesn't care, but she sure does pretty work

marcf