Biological validation of computationally determined gene-gene interaction

Question

How can a computationally determined three-way gene interaction be biologically validated? What kind of assays or tests must be performed using cell/tissue-based and/or mouse models to prove that the three genes may indeed have a joint effect?

Let's say, it is easier to identify and validate gene interactions that involve transcription factors like FOXM1. Consider a three-way gene interaction in Breast cancer - FOXM1-BUB1-CHEK1 - that can be tested for direct interactions via western blot and reporter assays. But such interactions may or may not be interesting - given the fact that transcription factors may be affecting expression levels of other genes. Most computational studies are focussed on identifying gene interactions based on coexpression or co-occurrence. There is literature on computationally identifying AND/OR relationships between interacting genes. I don't have a specific example to provide but, for argument sake, if we did suspect that three genes are interacting in an AND like manner, how do we biologically validate this finding? I'd also appreciate if you may have comments on the usefulness of such findings, particularly, with regard to designing more efficacious combination therapies against a disease".

Thanks

Answer 1

Hang in there, this answer will grow over time.

Just off the top of my head, I think we could write a small book (or a very long review article) to cover both the depth and the breadth of your question(s).

First, we need to define, and/or clarify, some terms. The term interaction probably means different things to different authors at different times. In fact, one author may use this term to mean different things at different times. So I would like to propose that we make some distinctions. For example, I would make a distinction between a genetic interaction between two genes (which is typically detected in one of the model organisms, using a genetic test between two loss-of-function alleles) versus a protein-protein interaction (sometimes called a PPI) between the two proteins encoded by those same two genes. PPIs have been detected classically by co-sedimentation in a sucrose density gradient, or co-immunoprecipitation (co-IP), or by affinity chromatography. More recently PPIs have been detected by mass spectrometry (MS) (sometimes coupled with affinity chromatography). PPIs have been inferred by using proxy assays, such as the yeast two hybrid (Y2H) assay.

PPIs can also be detected by surface plasmon resonance (SPR) assays, and there may well be others.

So, when you say gene-gene interactions, what precisely do you mean?

Answer 2

I'm gonna try to cover as many scenarios as I can think of:

• Gene interaction ( = co-expression):
  This is probably the easiest to validated (e.g. by western blot like you mentioned) but the hardest to interpret. Without further analysis of the genes functions you have no idea what effect their co-occurring activity might have. If the genes are metabolic enzymes or part of a signal cascade it would be a good idea to check both their individual roles in these pathways, but also check for potential synergistic effects. The same is in principal also true for transcription factors, but in that case 'pathways' are often not equally well understood, instead go for:

• Gene interaction ( = on same targets):
  In the case of transcription factors (but also e.g. Kinases/Phosphatases) that are co-expressed it might not unreasonable that they also affect the same targets. Here you need to again compare both individual effects (list of up/down regulated genes) with synergistic effects (e.g. geneA prevents normal function of geneB, geneC increases geneB function multifold). For transcription factor analysis you should generally analyse expression on the mRNA level (via qPCR) since the protein level introduces (at least) another layer of regulation.

• Protrein interactions:
  A good way to prove these experimentally is by using co-IP (immuno-precipitation) experiments: you use an antibody to retain a specific protein and all its (direct) binding partners. Then you show the presence of the partners you're interested in with a western blot.
  One caveat for this method with 3-way (or higher order) interactions is that - depending on the binding strength of the proteins - it can be very hard to prevent anything but direct interaction partners from being co-purified. If you want to absolutely prove that two proteins interact directly a yeast-2-hybrid (y2h) experiment is better suited.

Answer 3

Others have covered a lot more ground but I'll throw in a couple of approaches from the spectroscopy side since I've been looking at them recently.

Bioluminescent resonant energy transer (BRET) shows if proteins bind in vivo: "BRET measures the interaction of proteins using a bioluminescent donor fused to a protein of interest and a fluorescent receptor fused to its binding partner. The bioluminescent donor, usually a luciferase, does not excite the fluorophore using light, but transfers resonance energy through dipole-dipole coupling. To transfer resonance energy, the donor must be within 10nm of the receptor and in the proper orientation making the technique useful for measuring proteins in close proximity." (From https://www.promega.com/resources/pubhub/features/bret-nanoluc-luciferase-and-protein-protein-interactions/)

Surface plasmon resonance (SPR) also shows if proteins bind, but in vitro: "SPR is sensitive to changes in refractive index within about 150 nm from the sensor surface. To study the interaction between two binding partners, one partner is attached to the surface and the other is passed over the surface in a continuous flow of sample solution. The SPR response is directly proportional to the change in mass concentration close to the surface. Biacore systems can be used to study interactions involving (in principle) any kind of molecule, from organic drug candidates to proteins, nucleic acids, glycoproteins and even viruses and whole cells." (From the Biacore assay handbook)

These are both ways to look at direct interaction, ie binding. Functional interaction is a whole different story.