Topaz Train with simulate data

Hello everyone,

I’m currently working on training protein-specific particle pickers using the Topaz Train job, but I’m new to Topaz.

I have a good map from a previous dataset and am now working on the same complex using a new dataset. However, the map still exhibits some orientation bias, and the complex appears in various states within the dataset. I’m particularly interested in one specific state.

I’m considering generating simulated data from the good volume to train Topaz. I hope to use this trained model on the new dataset to avoid losing rare orientation particles.

I’ve added CTF and alignment information to the particles I selected from the dataset and applied them to the simulated data. However, the 2D class of the simulated data appear slightly different from those of the real particles. I’m concerned that this might lead Topaz to pick noise.

Best regards