Single Particle Analysis - "Gold Standard" version"

1. Background

The concept of the "gold standard" approach involves separating the particle image data into two mutually independent sets. These sets are then independently processed - i.e., the one set is never exposed to the information of the other, rendering any comparison apparently unbiased. Because the starting point of processing requires an initial reference map, and because particle picking is not without bias, the processing can never be completely without bias. The starting point should therefore be chosen with as little information as possible without making it practically impossible to get a result. The typical approach is to low-pass filter a map (to 30-60 Å) as the initial reference for both particle sets. In subsequent iterations, only the map reconstructed from the same set of particles in the previous iteration is used as reference. The expectation is then that if the processing is pushed to far into overfitting, that the resolution estimation based on a comparison of the two independent reconstructions will deteriorate.

SPA flowchart

The Bsoft workflow sets up two parameter files with alternating selection of particles. Each set is then independently aligned, selected and a reconstruction calculated. The reconstructions of the two independent sets are then compared by calculating the FSC curve. The amplitudes of the reconstructions are then weighed by the FSC curve to produce appropriate reference maps for the next iteration.


2. Preparation

The preparation for this workflow (CTF fitting and particle picking) is the same as for the typical SPA. The descriptions for orientation determination and reconstruction also apply, with the exceptions noted below.


3. Determining particle orientations

As with the normal SPA, each orientation determination run should be done in its own directory. In this directory, set up the input parameter file to designate the two particle sets:

bpartsel -verb 1 -all -alt 2 -out in.star ../run2/klh_run2.star

As with conventional SPA, the alignment can be done with borient to get global orientations, or with brefine to improve already established orientations.

a. borient:

For each of the input parameter files, launch borient:

borient -v 1 -select 1 -sym D5 -angle 2 -resol 25,150 -ann 10,90 -mode ccc -CTF -ref ../run2/klh_ref2_1.pif -out klh_run3_1.star in.star > klh_run3_1.log &

borient -v 1 -select 2 -sym D5 -angle 2 -resol 25,150 -ann 10,90 -mode ccc -CTF -ref ../run2/klh_ref2_2.pif -out klh_run3_2.star in.star > klh_run3_2.log &

b. brefine

For each of the input parameter files, launch brefine:

brefine -v 1 -select 1 -grid 2,0.2 -res 20 -mag 0.03 -defocus 0.002 -kernel 6,2 -ref ../run2/klh_run2_1.pif -out klh_run3_1.star in.star > klh_run3_1.log &

brefine -v 1 -select 2 -grid 2,0.2 -res 20 -mag 0.03 -defocus 0.002 -kernel 6,2 -ref ../run2/klh_run2_2.pif -out klh_run3_2.star in.star > klh_run3_2.log &


4. Particle selection

This should be done on each output parameter file:

bpartsel -v 7 -fom 0.34 -out klh_run3_1_sel.star klh_run3_1.star

bpartsel -v 7 -fom 0.34 -out klh_run3_2_sel.star klh_run3_2.star

The selection output files can be recombined ready for the next iteration:

bpartmulti -verb 7 -merge -output klh_run3_sel.star klh_run3_1_sel.star klh_run3_2_sel.star


5. Reconstruction

Two reconstructions should be calculated:

breconstruct -v 3 -resol 15 -rescale 0,1 -sym D5 -CTF baseflip -recon klh_run3_1.pif klh_run3_1_sel.star >& klh_run3_1_rec.log &

breconstruct -v 3 -resol 15 -rescale 0,1 -sym D5 -CTF baseflip -recon klh_run3_2.pif klh_run3_2_sel.star >& klh_run3_2_rec.log &


6. Resolution determination

This is the key step in the "gold standard" approach. The two independent maps are now compared to calculate an FSC curve and determine the resolution:

bresolve -v 7 -cut 0.3,70 -resol 15 -XML klh_run3_resol.xml -Post klh_run3_resol.ps -map klh_run3_1.pif klh_run3_2.pif


7. FSC filtering of reconstructions:

The FSC curve can be used to do an appropriate filtering of the reconstructions to more accurately represent the actual information content:

bampweigh -v 7 -FSC klh_run3_resol.xml -resolution 15 -rescale 0,1 klh_run3_1.pif klh_run3_1_fsc.pif`

bampweigh -v 7 -FSC klh_run3_resol.xml -resolution 15 -rescale 0,1 klh_run3_2.pif klh_run3_2_fsc.pif`