Howdy folks

I'm currently trying to create a reconstruction of a bacteriophage tail structure using the Relion helical reconstruction workflow. I manually picked ~150 tails or so and extracted segments containing 1 assymetric unit per box and got 3746 segments for 2d classing. I created 2d classes that looked quite good, and then used these for auto picking the original micrograph set; i used the same helical settings as the classing and extraction protocols, and used the write/read FOM maps features to optimise the picking threshold and inter-particle distance settings.

The problem I'm getting is that autopicking is picking helices in random noise, or on the sides of the grid, or in the heads of phage; worse still, it's very frequently not even picking the actual particles.

I tried picking more particles, and extracted and classed as 2 assymetric units per box as i figured perhaps a larger structure would be less prone to picking errors, however the error is the same. I tried playing with the maximum stddev noise parameter to reduce picking on the sides of grids but this appears to have achieved little. I also modulated the shrink parameter and reference low pass filter settings to see if that helped to no avail.

Iv'e attached a few images of my classes and an example of the weird picking.

Any ideas? I'm very close to just manually picking the whole set.

Cheers for any recommendations :)