Thread: "irreducible" complexity predicted in 1918

I'm starting to dislike comparing biological systems to machines.
With real machines, all you need to do is cut one of the wires and it's dead. It doesn't work like that with living systems.

Biological systems are very sloppy and noisy, this makes them robust and plastic. I don't think you will find many genetic changes that make the biological machinery completely break down.
E.g. humans can be born with extreme deformities, and still be alive.

The problem of "irreducible" complexity might have more to do with fitness landscapes. Once organisms got committed to a certain peak, it pretty much became a one way street.

I'm posting an image of embryonic development because this seems to be the best conserved pattern in "large scale" biology.



The developmental process isn't irreducibly complex, even though it's so conserved. Organisms start with the same development pattern and end up with very different body plan. It's just a consequence of its fitness landscape having very steep hills that are separated by huge gaps.

You won't see "irreducible" complexity at the lower levels either, even though they are extremely complex.

For example, this is the genetic network for Saccharomyces cerevisiae:


The topology of the network is basically the lowest system that determines how the whole organism will function.
Whole network can be analyzed by means of graph theory and only a few of all possible 3- and 4- node motifs will be found.

Uri Alon shows that these motifs function like logic gates and control systems(feedback and feed-forward mechanisms) and explains why only they are selected for.
They are easily added or lost by mutations of just a few bases in the regulatory regions of functional genes.

Adding or removing them from the network will usually not make the whole thing break down. Of course, deleting hubs from the network will completely fuck it up, but hubs with higher degrees are rare.

Other networks (protein and metabolic) are even more robust. This is because of their analogous nature, as opposed to mostly digital nature of the genetic system.

The information that I've been able to find so far on how to construct those graphs says:

"Each gene, each input, and each output is represented by a node in a directed graph in which there is an arrow from one node to another if and only if there is a causal link between the two nodes."

I'm a little confused. I guess we should stick to unicellular organisms. So an input is a state of the environment, an output is the production of a protein and a gene is the genetic material for a protein?

Is that right?

Also, will the graph be cyclic in general? That is, are there positive feedback loops or not?

Originally Posted by PhilosopherStoned

The information that I've been able to find so far on how to construct those graphs says:

"Each gene, each input, and each output is represented by a node in a directed graph in which there is an arrow from one node to another if and only if there is a causal link between the two nodes."

I'm a little confused. I guess we should stick to unicellular organisms. So an input is a state of the environment, an output is the production of a protein and a gene is the genetic material for a protein?

Is that right?

Also, will the graph be cyclic in general? That is, are there positive feedback loops or not?

That is a transcription network. It describes how products of certain genes influence expression of other genes.
Gene is defined as information needed for a functional protein (or non-coding RNA). In order to function, apart from the coding sequence, a gene must have other features as well.
One of these features is a promoter, a regulating sequence found upstream of the coding region.

Transcription factors (TFs) are proteins that bind to promoters and influence the transcription of those genes. They can either activate them or repress them. Some genes are also controlled by multiple TFs. Each TF has its own target sequence, but they sometimes overlap causing TFs to compete for binding (competing TFs would have opposite effects on that gene)

Interesting to note: around 10% of all human genes are thought to be TFs

Most TFs are, abstractly speaking, intracellular representations of extracellular conditions.
E.g. nutrient concentration drops, TFs that regulate starvation response become active and turn on (and turn off) their respective gene targets.

(discovery of this mechanism earned Jacques Monod a Nobel prize, google "Lac operon" if you're interested in more details)

In the graph nodes represent genes and edges represent interaction.
X------>Y means: "the product of gene X is a transcription factor for gene Y"
Hubs are genes that code for TFs involved activation and repression of many proteins.

To answer you other question: out of all possible 3-node motifs, the positive feedback loop is the only one that's significantly overrepresented in transcription networks. It can function as AND or an OR logic gate, depending on other parameters.

Also, a lot of genes regulate their own expression. I'm not sure, is a node connected to itself a cyclic subgraph?

Originally Posted by SnakeCharmer

Also, a lot of genes regulate their own expression. I'm not sure, is a node connected to itself a cyclic subgraph?

Yes. Thanks for the info. This is very interesting stuff. I've only ever studied evolutionary biology from the perspective where a gene is defined to be "the genetic matter which codes for a heritable phenotypic trait." That definition gets used by ethologists I suppose. I've been getting curious about going deeper recently though.