Illuminating the Interactome

A massive screen yields the most comprehensive map of binary human protein interactions to date.

By Molly Sharlach | November 20, 2014

A systematic map of nearly 14,000 binary interactions among 4,100 human proteinsTHOMAS ROLLAND

The completion of the human genome sequence more than a decade ago was an indisputable triumph for biomedical research. And more recently, efforts such as the Encyclopedia of DNA Elements (ENCODE) project have sought to expand knowledge of functional elements within the genome. 

But truly connecting genotype to phenotype will require a comprehensive view of how the protein products of genes operate and interact. Researchers at the Dana-Farber Cancer Institute’s Center for Cancer Systems Biology and their colleagues have produced a new human interactome map, reported today (November 20) in Cell. The map is based on a systematic screen of 13,000 human proteins that uncovered 14,000 pairwise interactions.

This nine-year project likely represents about 5 percent to 10 percent of all the protein-protein interactions that exist, according to study coauthor Fritz Roth of the University of Toronto. While still limited in scope, it is at least a five-fold improvement over previous interactome maps, Roth added.

“This is a long road, and we’ve never had a human interactome project to go with the Human Genome Project,” he said. “But I think people are starting to appreciate that the genome is the beginning of the story . . . it’s a parts list in an alien language that we’re starting to figure out.”

To identify these interactions, the researchers used a high-throughput yeast two-hybrid approach, in which 82 million protein pairs were each tested four times in two different configurations for their ability to activate a reporter gene in yeast. The researchers also validated selected interactions using three independent methods, testing whether the protein pairs could reconstitute the parts of a fluorescent protein or a membrane-bound protein complex in mammalian cells, plus used an in vitro method.

The team compared its results to a list of protein interactions supported by multiple pieces of evidence garnered from a literature search in 2013. The researchers found that their systematic strategy picked up a large swath of interactions that were missed by individual studies.

“This kind of centralized approach has a much higher likelihood of finding interactions throughout the human proteome, rather than just finding interactions of the specific proteins that people have studied because of a disease process or because of the specific cellular function that they’re interested in,” said Stanley Fields of the University of Washington. Fields, who pioneered the use of the yeast two-hybrid system, was not involved in the present study, but served on the advisory board for a National Institutes of Health (NIH) grant that partially funded the research.

Notably, the new interactome map lends support to a long-held suspicion that proteins implicated in cancer participate in disproportionate numbers of interactions with other proteins. As scientists sequence tumor genomes, more extensive knowledge of these protein-protein interactions could help to distinguish “driver” mutations that cause cancer from “passenger” mutations that are simply along for the ride.

“Our goal is to help facilitate the expansion and robustness of the human interactome to the point that it can really provide insight into every chronic disease,” said Joseph Loscalzo of Harvard Medical School, a cardiovascular researcher and long-time proponent of “network medicine” who has collaborated with study coauthor Albert-László Barabási, but was not involved in the present work.

While this latest map is a valuable resource, it provides a static view of the proteome, said Loscalzo: “Looking at dynamic changes will be another important part of this . . . it would also be useful to look at adaptive responses of the proteome to stresses in the environment.”

Roth and his collaborators are already at work on the next interactome map, which will expand the screen to 17,000 proteins. While large, this expanded map will still be far from comprehensive. “One thing we know is that not every interaction can be detected by every assay, so it’s unlike genome sequencing,” he said. “It’s an asymptotic problem.”

Fields suggested that future efforts might consider tissue-specific proteins and posttranslational modifications that affect protein-protein interactions, in addition to proteins produced from alternatively spliced transcripts. “Ultimately, the interaction maps are going to be way more complicated than just the genomic sequence has proven to be,” he said, “but you have to start somewhere.”

T. Rolland et al., “A proteome-scale map of the human interactome network,” Cell, 159:1212-26, 2014.

Add a Comment

Avatar of: You



Sign In with your LabX Media Group Passport to leave a comment

Not a member? Register Now!

LabX Media Group Passport Logo


Avatar of: James V. Kohl

James V. Kohl

Posts: 516

November 21, 2014

"...you have to start somewhere.”

In our 1996 Hormones and Behavior review, we started with RNA-mediated cell type differentiation in yeasts and linked them to sex differences in cell types in mammals via nutrient-uptake, DNA methylation, and the metabolism of nutrients to species-specific pheromones that control the physiology of reproduction.

Excess nutrient stress and social stress appear to be linked to negative outcomes via the bio-physically constrained chemistry of protein folding, which has since been linked across species to the differentiation of all cell types in all individuals via the conserved molecular mechanisms of alternative splicings reported in our section on molecular epigenetics.

Proteome-scale maps that include information about the amino acid substitutions that differentiate cell types are the next step forward in the context of RNA-mediated events. The RNA-mediated events can be distinguished from evolutionary events, whicht have not been substantiated with any experimental evidence that evolutionary events actually occur.

Instead, ecological variation appears to lead to ecological adaptations in the absence of stress-induced accumulation of deleterious mutations, not by natural selection for anything besides food odors and pheromones, which control the physiology of nutrient-dependent reproduction in my model.

As we now see, starting from the perspective of mutation-driven evolution has delayed focus on RNA-mediated events that link the epigenetic landscape to the physical landscape of DNA in the organized genomes of species from yeasts to primates via amino acid substitutions that differentiate their cell types via protein-protein interactions.

Avatar of: standishman


Posts: 5

November 21, 2014

Do all these interactions lend support to Keith Porter's microtrabecular lattice view of cytoplasmic organization?

Avatar of: James V. Kohl

James V. Kohl

Posts: 516

Replied to a comment from standishman made on November 21, 2014

November 21, 2014

Thanks for asking, but that's not a question that can be answered with a yes or no, without comparing his view to the model of biologically-based cause and effect published as Nutrient-dependent/pheromone-controlled adaptive evolution: a model. Please feel free to ask questions about the comparisons you want me to specifically address.

The model was developed in accord with what we stated about  the molecular epigenetics of cell type differentiation in From Fertilization to Adult Sexual Behavior

Excerpt: "Yet another kind of epigenetic imprinting occurs in species as diverse as yeast, Drosophila, mice, and humans and is based upon small DNA-binding proteins called “chromo domain” proteins, e.g., polycomb. These proteins affect chromatin structure, often in telomeric regions, and thereby affect transcription and silencing of various genes (Saunders, Chue, Goebl, Craig, Clark, Powers, Eissenberg, Elgin, Rothfield, and Earnshaw, 1993; Singh, Miller, Pearce, Kothary, Burton, Paro, James, and Gaunt, 1991; Trofatter, Long, Murrell, Stotler, Gusella, and Buckler, 1995). Small intranuclear proteins also participate in generating alternative splicing techniques of pre-mRNA and, by this mechanism, contribute to sexual differentiation in at least two species, Drosophila melanogaster and Caenorhabditis elegans (Adler and Hajduk, 1994; de Bono, Zarkower, and Hodgkin, 1995; Ge, Zuo, and Manley, 1991; Green, 1991; Parkhurst and Meneely, 1994; Wilkins, 1995; Wolfner, 1988). That similar proteins perform functions in humans suggests the possibility that some human sex differences may arise from alternative splicings of otherwise identical genes."

Alernative splicings have since been shown to be the nutrient-dependent drivers of all cell type differentiation in all invertebrates and vertebrates, which also links them to pheromone-controlled cell type differentiation in microbes via the physiology of species-specific reproduction. An extension of my model of ecological speciation to marine mammals can be found in "The differences in amino acid composition among different tissues can lead to large differences in trophic discrimination 38."

Dobzhansky (1973) may have been the first to note that: "...the so-called alpha chains of hemoglobin have identical sequences of amino acids in man and the chimpanzee, but they differ in a single amino acid (out of 141) in the gorilla."

Avatar of: James V. Kohl

James V. Kohl

Posts: 516

Replied to a comment from James V. Kohl made on November 21, 2014

November 21, 2014

Re: Keith Porter. Also see: Chromatin measurements reveal contributions of synthesis and decay to steady?state mRNA levels It appears to place RNA-mediated cell type differentiation via amino acid subsitutions into the context of thermodynamic cycles of protein biosynthesis and degradation via links to genomic stability that arises in the context of the amino acids substitutions, NOT MUTATIONS.

"Regulation of mRNA levels is a key mechanism that defines cell identity. Cellular homeostasis requires stable gene expression patterns, while differentiation events in metazoan development or responses to external stimuli involve resetting of the transcriptional program. During the lifespan of an mRNA from transcription over maturation, export, translation, and decay, its activity and abundance is controlled by various mechanisms: histone modifications and DNA methylation determine the epigenetic state of the chromatin environment of a gene depending on the DNA accessibility the transcription machinery can bind and initiate transcription and thereby produce primary transcript at different rates (Segal and Widom, 2009; Bell et al, 2010). This is modulated co?transcriptionally by splicing and poly?adenylation (Millevoi and Vagner, 2010; Nilsen and Graveley, 2010; Di Giammartino et al, 2011) and further regulated at the level of nuclear export. Once the mRNA is in the cytoplasm it is subject to further post?transcriptional processing, which can reduce the transcript level in a targeted manner. Two major post?transcriptional regulatory processes influencing the amount of mRNA molecules available for translation are general RNA decay and microRNA?mediated RNA interference."

Avatar of: James V. Kohl

James V. Kohl

Posts: 516

Replied to a comment from James V. Kohl made on November 21, 2014

November 21, 2014

I just realized the the senior author of the latest article is also a co-author on the open access article: Protein interaction network of alternatively spliced isoforms from brain links genetic risk factors for autism.

I've requested a reprint of the latest work, but based on other publications expect it to continue to support my insistence that nutrient-dependent cell type differentiation via amino acid substitutions be considered in the context of health for comparison to mutated proteins that should be considered in the context of accumulations linked to disease.

Evolutionary theorists like to think that mutations are beneficial to increasing organismal complexity, but won't tell us how that is possible outside the context of pseudoscientific nonsense associated with population genetics.

Avatar of: C.Wyman


Posts: 1

November 24, 2014

I question wether this sort of study will illuminate biological function.  Possible any two proteins will "interact" in the right circumstances.  The binary yes or no "interaction" is not so informative.  Without knowing more; the strength of the interaction, the durration, the consequences.... it is not clear to me that this map is useful.  Large numbers are impressive but are not an end in themselves.  I also argue that the illustration of the interactions in the "hairball" display is not at all informative.  If this is useful information a manner to present it in a way that enhances understanding and usefulness needs to be developed.

Avatar of: James V. Kohl

James V. Kohl

Posts: 516

Replied to a comment from C.Wyman made on November 24, 2014

November 27, 2014

Re: I question wether this sort of study will illuminate biological function.

Were you taught to believe that mutations and/or natural selection somehow link the epigenetic landscape to physical landscape of DNA in the organized genomes of species from microbes to man?

If not, you may recognize the fact that others report research results in the context of what is currently known about the bio-physically constrained chemistry of protein folding. For example, serious scientists now incorporate what is known about RNA-mediated events and amino acid substitutions that differentiate all cell types in all individuals of all species via conserved molecular mechanisms.

See: Clinically Actionable Genotypes Among 10,000 Patients With Preemptive Pharmacogenomic Testing

If you are an evolutionary theorist, please consider learning enough about physics, chemistry, and biology to become a serious scientist.

Other researchers who understand the molecular mechanisms of epigenetic pharmacology and personalized medicine are waiting for your help.

You, too, can contribute to the understanding of how metabolic networks and genetic networks are linked. Clearly, it is time to frame research results in levels of evidence, not in the ridiculous theories invented by population geneticists.

Avatar of: Subhash


Posts: 6

November 27, 2014

Two thing: - looking at the interactions as they are presented in a diagram gives one the feeling that we understand the cell quite well because we can fill in thoussnds of molecules and join them with arrows - we know that it is quite far from the truth. Another thing it brings out is the discussion of the 'predestined' nature of deveopment which is largely true at the organism level with superimposed 'free will' poosibilities of cells taking paths that they may choose later - sometime to the detriment of the organism. Then there are the stem cells, whose paths we eant to experimentally modify - So is life!

Avatar of: Oomemes


Posts: 2

Replied to a comment from C.Wyman made on November 24, 2014

December 1, 2014

Please read the actual paper to find:

-evidence that not just any protein pair interacts in this assay

-evidence of biological relevance of the detected interactions

-no hairball pictures whatsoever

Avatar of: PatrikD


Posts: 7

Replied to a comment from C.Wyman made on November 24, 2014

December 1, 2014

Quote C.Wyman

I question wether this sort of study will illuminate biological function.  Possible any two proteins will "interact" in the right circumstances. 

That is not what the data suggests. As Roth mentions in the article, "this is an asymptotic problem." That is, as more studies like this are being done, it is clear that the total number of interactions is reaching some sort of asymptote - which is NOT "all proteins interact with all proteins".

Note that this study only found 14,000 pairwise interactions, out of 82 million pairs of protein tested. That is only a very tiny fraction of protein that do actually interact - about one in 6000 pairs of proteins tested.

The authors also provide an estimate that this is likely to be 5-10% of all the human protein-protein interactions. In other words, if they extrapolate the asymptote, that is where the number of real interactions starts to saturate. So they predict 140,000-280,000 interactions, or an average of around 7-14 interaction partners per protein (including splice variants).

As for the hairball - totally agree with you there. It provides no useful information other than "it's complicated!" But it's a sexy figure, so it tends to get reused over and over at the expense of more sophisticated and information-rich representations that unfortunately carry a lot less impact because they take a little more effort to explain.

Popular Now

  1. Estonia Offers Free Genetic Testing to Residents
  2. Human Brain Organoids Thrive in Mouse Brains
  3. RNA Injection Restores Hearing in Guinea Pigs
  4. Jim Bridenstine Confirmed to Lead NASA