Issues with Pocket Volume Calculation and Snapshot Filtering in mdpocket

[ArseN4272]

Hello,

I am using mdpocket for filtering snapshots based on pocket volume, and I’ve encountered two main issues I would like to address:

Multiple pockets in snapshots:
When I use an iso file as a reference (with one specific pocket selected), I still get many snapshots where multiple pockets are detected. The descriptor file seems to only report the total pocket volume of the snapshot, but I need a way to filter snapshots based on individual pocket volumes rather than the total. Is there any way to extract the volume of each pocket separately for better filtering?

Volume discrepancy between mdpout_mdpocket.pdb and descriptor file:
I noticed a discrepancy between the volume calculated using Monte Carlo simulations (based on the pocket's sphere radius from the B-factor column in mdpout_mdpocket.pdb) and the volume listed in the descriptors file. For example:

Using the sphere radius and Monte Carlo method, I get a volume of 228.

In the descriptors file, the volume is 41.

Even when using a snapshot with a single pocket (where I can apply the volume formula for a sphere:
4/3πr^3, the calculated volume is 220, while the descriptor shows 44.6.

This significant discrepancy leads me to suspect that either:

I’m missing some crucial information or volume metrics when calculating the pocket volume.

Or there might be some other metric being applied in the descriptor file that I’m unaware of.

Even when I tried changing the value of the -v parameter from 2500 to 1,000,000, but it did not seem to change the results much.

Can someone explain how the volume is calculated in the descriptor file, or if I’m misinterpreting the B-factor values in mdpout_mdpocket.pdb?

Thank you in advance for your help!

[pschmidtke]

You can have multiple alpha spheres overlapping in 3D space. The sampling algorithm for volume calculation places grid points in a box around an area of interest and samples whether the point is inside or outside any sphere. If inside then the volume of the voxel is accounted for, if outside not.

[ArseN4272]

Thanks for the explanation! I understand the overlap-based voxel sampling method now.
However, I’d like to clarify one detail I might not have stressed enough in my original post:

In one test case, I used a pocket that was described by a single alpha sphere, so there shouldn’t be any overlaps involved. Even in this minimal setup, the volume calculated via 4/3​​πr^3 using the B-factor radius gives me ~220 ų, while the descriptor file shows only ~44 ų.

Is there any reason such a large discrepancy would exist even for a single-sphere pocket?

Thanks again for your time!