Spotting differences in large structure collections
The context
The amount of structural information that we’ll have to cope with in near future will become overwhelming given our current ways to work on complex 3D information. Especially streamlining the analysis of larger sets of structures is still a challenge using standard visualization and molecule modelling tools.
One of the insights one could get analyzing larger collections of structures is how the protein moves / adapts or changes in presence of different conditions (different soaked ligand for instance). How a mutation affects protein conformation / dynamics.
In the end you generally want to be able to spot, what is the common thing between all of my structures and where are the differences?
Our friends at Abbvie requested such a feature for quite some time in 3decision, mainly to detect differences in the binding site very quickly (mutations & movements).
Spotting local differences quickly
On the left: chek1 kinase structure (5dls) superimposed to chek2 kinase structure (4bdc) with the default binding-site focused representation. On the right, the focused “difference” mode. All backbone is shown as thin ribbon, unless a major conformational change (or multiple mutations) occurs. All mutations or movements on the amino-acid level are shown as sticks.
Spotting quickly the differences in the ATP binding site, as shown above, allows you to highlight main differences here such as:
mutations between structures
unresolved residues
larger conformational changes (side-chain or backbone)
Especially when working with very similar homologs or when you are browsing structures for the first time on a new project or just got in a novel structure from your crystallography team, this can get really handy.
Spotting global differences quickly
What we could see above was focused on a particular location. That’s mainly to help molecular modellers, chemists and crystallographers to focus on an area of interest and to declutter the view compared to a classical full-structure view.
However, we also integrated a way to show differences on the full structure. That’s definitely not useful if you have large movements. However, if you want to quickly analyze for instance fragment screening results from initiatives like PanDDA from XChem such a feature can become quite handy.
Example of showing all the differences on a larger ensemble of structures. All “major” movements get shown as lines or cartoon representation. The binding site is very stable and movements on the protein surface can be observed mainly on two locations.
The tool, available in the 3D Viewer toolbar, also allows you to adjust the RMSD value you’d like to consider for a residue to be considered as different when it’s moving. This essentially enables the user to make the selection more or less sensitive.
Last, rather than focusing only on differences, you can also choose to invert that and show only the common parts between two structures. Again this can help to clarify visually several areas of interest in your structure and streamline visual inspection, rather than going over each portion one by one.
Conformation sampling
Much has been written and discussed about potential applications of algorithms like alphaFold, enabling automated modelling or conformational sampling using molecular dynamics. The outcome of any of these techniques are conformations of a biomolecule. What is to be expected is the increase in the output of such in-silico but also experimental structures, so finding a way to streamlining their analysis could become a valuable feature in the foreseeable future.
How can I use that?
This feature is available in the new 3decision release, about to be released this week. If you already have access to 3decision, check out the little light bulb icon in the 3D Viewer toolbar. If you do not have access to 3decision yet, feel free to contact us here.