I’ve learned from another post that 100-500 particles are recommended for each reconstruction used in the PCA initialization mode. I was wondering if there’s a suggested lower bound for the number of particles to use before noise starts to overwhelm the reconstructed volume and skew the PCA. Can I go below 100, say to use 50 or even 30 particles for each reconstruction? Only about 5% of my particles have the target protein bound (I’ve optimized the sample prep and this is the best I can do). If I use 500 particles, the density from the 5% bound protein will get greatly diminished by the 95% unbound particles, but if I use just 30 particles, some of the reconstructed volumes may happen to come from a much higher percentage of the bound particles due to complete randomness. Would this be a sensible thing to do?
Honestly I wouldn’t bother using the PCA initialization mode - I haven’t found it to make much difference under any circumstances.
As an alternative, you could initialize using volumes from 3D-VA display (generated in “intermediates” mode where it calculates reconstructions along each component
Thanks for your suggestion! I’ll definitely give that a shot.