VESPER uses a combination of mean shift algorithm and fast Fourier transform (FFT) to identify
the best superimposition of two EM maps. Source codes are available
here. Alignment using conventional Cross-correlation and the Laplacian filter is available in the provided code.
(Nov. 8. 2021) Bug fix:
A major bug on "Step1: Search parameters" is fixed. Those parameters will work properly.
VESPER has two major steps:
Identify the best superimposition between two EM maps with certain voxel spacing and angle spacing.
Calculate the normalized z-score for each of top 10 models in VESPER output. Given a query map,
normalized z-score is the score we use for ranking all other maps in the database. Empirically, related
maps have normalized z-score larger than 10.0.
Search results by VESPER
In Step 1, choose the voxel spacing (v, in Å) and angle spacing (A, in degrees). VESPER uses v = 7 Å and A = 30° by default,
which takes 1 hour on average. Usage of smaller voxel spacing and angle spacing can improve the accuracy
in fitting, but also makes the calculation much slower.
In Step 2, provide a reference map ID and a target map ID or upload two EM map files. The reference map is fixed. The target
map is superimposed onto the reference map.
In Step 3, provide the Email address. After the calculation is done, VESPER sends the calculation results to the provided Email address. To visualize alignment results, follow the instruction included in the zipped files.