Evolution
Intelligent Design
The Interactome Multiplies Specified Complexity
One aspect of the “Unknome” is starting to become clearer: the interactome. If the Unknome refers to the set of components in a cell we know nothing about, the Interactome refers to “the whole set of molecular interactions in a particular cell.” A recent paper has created a “wow moment” about the interactome. It found that there are far more interactions between proteins than previously thought.
A New Method
Publishing in Nature, seven researchers from Germany and Denmark explored the “social and structural architecture” of proteins in a eukaryotic organism. Michaelis et al. created a new method for investigating interactions between proteins. The interactome, they say, has been studied for two decades, but work has been tediously slow due to procedural challenges.
The large-scale study of cellular interactomes using mass spectrometry-based proteomics dates back over 20 years, culminating in 2 studies in which nearly half the expressed yeast proteome was successfully purified with identified interactors. These datasets have been mined extensively, leading to a network-based view of the cellular proteome. Given the importance of the interactome for functional understanding and the substantial improvements in mass spectrometry technology during the past decade, we set out to generate a substantially complete interactome of all proteins present in an organism in a given state. We made use of an endogenously GFP-tagged yeast library containing the 4,159 proteins that are detectable by fluorescence under standard growth conditions. [Emphasis added.]
They used a refined “pull-down” method to “bait” known proteins tagged with green fluorescent protein (GFP) and then observe what other “prey” proteins connected to them.
Miniaturization and standardization of the workflow in combination with an ultra-robust liquid chromatography system with minimal overhead time coupled to a sensitive trapped ion mobility mass spectrometer utilizing the PASEF scan mode resulted in very high data completeness across pull-downs. This workflow required only 1.5 ml instead of litres of yeast culture, provided a constant throughput of 60 pull-downs per day and enabled the use of the same conditions for soluble or membrane proteins of vastly different abundances (Fig. 1a).
The new work doubles the number of proteins studied and triples the number of interactions found “compared with existing interactome maps.” They checked and cross-checked the data for accuracy. The results were startling.
The replicate GFP pull-down measurement in the 4,147 yeast strains resulted in the enrichment of 82% of the baits (Extended Data Fig. 1). Our mass spectrometry data provided statistically significant evidence for more than 30,000 physical interactions, corresponding to an average of 15.8 interactions per protein. Most were supported by forward pull-down (35%), followed by forward pull-down and significant prey correlation (29%), whereas nearly all interactions with both forward and reverse evidence also had significant correlation z-scores (95%) (Extended Data Fig. 2).
More than two-thirds of the interactions discovered were novel, they said, not previously reported. While a small percentage of the baits did not retrieve “prey” proteins, that doesn’t mean they do not interact.
Altogether, based on the total of 4,403 identified yeast proteins, with 74.1% having at least two interactors, 15.1% had one and only 10.8% had no discernable interaction partner. To investigate whether the latter set is truly ‘non-social’ or is an artefact of expression level or its tag position, we performed our workflow on a subset of the proteins using N-terminal tagged strains with identical promoters (Extended Data Fig. 5). This yielded additional interactors for about half of the proteins. Notably, the overall average of identified interactors in this set was around 2, compared with 16 in the main dataset, indicating that this set of proteins was indeed poorly connected (Supplementary Fig. 2). Although reciprocal tagging was beneficial, complexes with higher numbers of interactions would already be picked up by the redundancy effect of our screen. Given that some of our baits will have context-dependent interactions that are not captured here, our estimates are conservative and we conclude that almost all yeast proteins are ‘social’.
Remember, This Is Just Yeast
Keep in mind that all these 30,000+ interactions between 4,159 proteins are taking place in yeast — the smallest and simplest of eukaryotes! One can only imagine the enormous number of interactions taking place in the cells of higher organisms possessing tens of thousands of proteins. In complex multicellular organisms like us, furthermore, interactions extend upward into additional dimensions: between cells, between tissues, between organs, and between organisms.
This nearly saturated interactome reveals that the vast majority of yeast proteins are highly connected, with an average of 16 interactors. Similar to social networks between humans, the average shortest distance between proteins is 4.2 interactions.
The findings from Michaelis et al. blow the lid off any notion of “simple” cells. Stationary diagrams of cells tend to depict the parts as loners: a mitochondrion here, a ribosome there, a vacuole over yonder. This work shows that the parts are in a buzzing hive of activity, with everything communicating, touching, releasing, migrating, and reconnecting. By analogy, think of a still picture of a city compared to a time-lapse video of the scene, with cars and people moving about in a multitude of ways to talk, work and accomplish individual and collective goals.
The Social Network
This paper also blows the lid off notions of cellular “junk.” If so-called “junk DNA” were generating “junk proteins,” much of the cell would be like hordes of the jobless on the streets taking up space and wasting resources. Instead, these proteins all have places to go and things to do. Everyone is contributing to the success of the social network. The unemployment rate in a cell is so low, it may not even be measurable. “The high connectivity of most proteins organizes almost all of them (3,839) into a single giant connected component,” the authors state, “accompanied by 41 small components (88 proteins)” acting, we might portray, like subcontractors.
If so, there are no unemployed proteins. The situation recalls to mind the ENCODE project that found over 80 percent of the genome was transcribed. And the closer they looked, the more they found function in what was considered genetic junk.
The Design Inference
The interactome can be added to the huge list of biological phenomena exhibiting the two requirements for the design inference: specification and low probability. Explained in the newly expanded and revised edition of The Design Inference by William Dembski and Winston Ewert, those two qualities in every phenomenon — as evidenced by each case in which we have access to its history — rule out chance and natural law, leaving intelligent design as the inference to the best explanation. The “interactome” in a large company making jets or cars, for instance, would never come about by the law of electrodynamics or by random groups of people finding themselves in the same building. The purpose preceded the parts and actors.
Critics of ID try to carve out biology as a special case due to the presumed stepwise gains of natural selection. In the Introduction their book, Dembski and Ewert face the claim that natural selection is a designer substitute, a blind watchmaker that can climb Mount Improbable.
Since the publication of the first edition of this book, the debate over the design inference and its applicability to evolution has centered on whether such gradual winding paths exist and how their existence or non-existence would affect the probabilities by which Darwinian processes could originate living forms. Design theorists have identified a variety of biological systems that resist Darwinian explanations and argued that the probability of such systems evolving by Darwinian means is vanishingly small. They thus conclude that these systems are effectively unevolvable by Darwinian means and that their existence warrants a design inference. In this book, we recap that debate and contend that intelligent design has the stronger argument.
The interactome adds more real-world evidence for the stronger argument.