Methods to analyse structural similarity: co-evolution vs convergence

Question

[kelpotus22]

Question: Some studies report structural mimicry/similarity between HIV proteins or their host receptor (gp120, CD4, CCR5, CXCR4) and certain venom toxins. I also came across a study suggesting co-evolution of venom-related genes with immune/self-resistance genes. Could the HIV–toxin similarity be a case of co-evolution (or convergent evolution)?

What I would like to be advised on: So if my hypothesis is that : venom toxins and HIV proteins may share convergent structural features that mimic host receptor interfaces. Now I want to explore this from a molecular perspective (using available genomics/proteomic/structural data from UniProt, PDB, etc.). What computational approaches would be best suited to test whether this is true co-evolution versus structural convergence?

I appreciate any insights you all might have.

[dthorbur]

A few comments.

1. Coevolution of a trait between 2 organisms can end up with a convergent structure with another organism. It's not either/or.
2. Please show some research. We respond better when you have an idea, and want to refine it rather than asking us to design an experiment for you. There are many different ways you could start this project. Many ideas can be found in published papers with similar questions.

Answer 1

[Joe]

Without historical data (ancient DNA) there's no way to ever really know this for certain. The best we can do is make inferences based on the 'distance' between taxa, so there is actually no real way to categorically say "these 2 genes are true orthologues" versus just convergent evolution.

That said, it would be generally accepted that if you can show two proteins have very similar 3D structure, and almost no sequence similarity (particularly if they originate from specied far apart on the tree of life), you'd have a strong argument for saying its convergent evolution.

Broadly speaking, I would tackle this in two ways potentially, though there might be a more robust approach I'm not aware of:

1. Do sequence-based informatics analyses (demonstrate that the nucleotide/protein identity is very low, and that these sequences diverged X years ago based on phylogenetic inference).
2. Do structural comparisons such as docking, superposition and deviation calculations.

If you could arrive at a position that said something to the effect of "these proteins are only 5% similar at the amino acid level, but have an relative RMSD score for their active sites of <3 Angstroms" (picking numbers at random), they therefore must have evolved to address similar structural constraints.