search

zhonge / cryodrgn_empiar

24 stars 3 forks source link

Training and analysis based on simulated datasets #3

Open qiliu-ghddi opened 2 years ago

qiliu-ghddi commented 2 years ago

@zhonge

3 simulated datasets (simulated.uniform, simulated.ribosomes, simulated.cooperative) and 50 density maps (simulated.models) are available on Zenodo.

We can reimplement the result quite well on the simulated.ribosomes dataset.

cryodrgn train_vae combined.mrcs --ctf ctf.pkl --poses poses.pkl --zdim 8 -n 50 -o ribosomes_combined_outputs
cryodrgn analyze ribosomes_combined_sz128 49 --flip --Apix 3.0

But we cannot reimplement the results of the other two datasets as good as that in the paper. The training and analysis commands are following:

# simulated.cooperative
cryodrgn train_vae particles.128.ctf.mrcs --ctf ctf.pkl --poses poses.pkl --zdim 8 -n 50 -o cooperative_sz128
cryodrgn analyze cooperative_sz128 49 --flip --Apix 6.0

# simulated.cooperative
cryodrgn train_vae particles.128.ctf.mrcs --ctf ctf.pkl --poses poses.pkl -zdim 8 -n 50 -o uniform_sz128
cryodrgn analyze uniform_sz128 49 --flip --Apix 6.0

It will be a great help for us if the right commands or parameters are available.

Thanks!