Using 3DVA to calculate the distribution of each state

Hi, all

In my case, I want to use the cluster mode of 3DVA to classify the different conformations of my protein and also analyze the population of each conformation.
After doing ab-initio (5 classes) and heterogeneous refinement, I could only get one full reconstruction and others were incomplete reconstructions of the target protein. The incomplete reconstructions are also part of my protein, not composed of junk particles.
So, my question is that should those particles be included for 3DVA ?
If I only use the particles of the full reconstruction, the population of different conformations might have some bias?

And in order to generate a mask for 3DVA, I will continue to do a NU refinement after heterogeneous refinement.
If I want to use all of the particles for 3DVA, should I do the NU refinement with all the particles and the volume of the full reconstruction?
What is the correct way to analyze the population of each conformation?
Any suggestion?



Hi Jessica, sometimes it can be tricky to understand the particle diversity within the dataset, especially when you come across “partial reconstructions” that contain only part of the total protein. This could be due to heterogeneity in the region missing, in which case you would want to include these particles for downstream characterization/separation. Alternatively, it could be that the missing region is denatured and that those particles should be excluded. See this study by D’Imprima.

I think the best way to address concerns like this is to perform refinements and 3DVA on the total population of particles, and also on the subsets that are separated with your method of choice. Looking at a single refined structure is somewhat informative these days but further classification and/or 3DVA is the most helpful way to understand what is really going on and how to address it. You may start out undertaking multiple processing pathways to understand the data (ie different ways to separate) and then go back and reprocess it with a more focused strategy when you know what works and where to look.

I would start with a NU refinement of all the particles and subsequent 3DVA using a mask that includes regions that appear heterogeneous (refine mask may be ok, but be skeptical of the fsc mask as it often masks out moving regions).

Try running 3DVA or heterogeneous refinement on the particle set corresponding to the incomplete reconstructions. Also run it on those leading to the full reconstruction. Sorry if this is vague and inconclusive. Basically, try lots of things, see what comes out, repeat.



Thanks very much for your helpful reply.
I’ll try to do 3DVA with both datasets that includes all of the particles and that includes only the complete reconstruction. I started to think the partial reconstruction is from denatured particles since it only includes the N-ter part of my protein, and the part is where the flexibility comes from.

It’s hard to know for sure what’s going on sometimes. Once I start to see the same thing by different processing methods I start to believe it’s real.