Can the results of 3DVA be totally unaffected when using downsampled particles?


I have to use the dowsampled (at least 1.5 X binned) particles to run the 3DVA jobs since the dataset is too large.
And I am wondering if using the downsamples particles will give different results?

Generally, I use both simple mode and cluster modes to output movie series and volumes, respectively.
For volumes, I do the NU refinement with the original particles. But I am worrying if the particle distribution of each cluster would be influenced due to using downsampled particles.

For movie series, the output is using downsampled particles. Can it be replaced by the original particle set?


Hi Jessica,

I’ve always used downsampled particles and 3DVA runs insanely fast when particles are downsampled. That said, there are a couple of places where you can restrict the resolution. I have used 3DVA to look for large scale (>20A) movement of domains. 3DVA is advertised to be able to resolve 3-4A differences. Now, depending on what resolution cutoffs you set, you will or will not be able to see these differences. So, to answer your question, it’ll depend on the range of motion you are looking for. My strategy has been to downsample (considerably, 2-4x) and then look for the range of motions. Then I reextract these particles that have been separated (usually using the cluster mode) at the required higher box size and do my refinements. This has worked really well and I have been able to separate out the distinct conformations. I would urge you to look at the resolution cutoffs and play with them to see if you get similar results with and without downsampling. Hope this helps.


1 Like

Hi, Vamsee

Thanks for your sharing and helpful advice.
I am wondering how do you re-extract those particles at specifically larger box size?
I use the “blob” particle dataset before downsample to replace the downsampled one in a refinement job.
Is there another way to do this?


Yeah, that’ll work if you have extracted the particles at the appropriate box size already. If not, you could just take the downsampled particles and run an extract micrographs job just without the fourier cropping, at the original box size. Either way you should get almost the same number of particles.

Got it.
Thank you very much. :blush:

1 Like