My theozyme looks very disorganized after the CA_diffusion and refinement steps. The rigidity is not maintained at all and there seems to be overlap of atoms in some cases. I could provide the input script I am using in both cases. I have changed the values of inference.refine_recycles in the range of 4 to 20, diffuser.T from 50 to 15, denoiser.noise_scale_frame from 0.05 to 0.01. none of these measures seem to give any useful results. Please help in this regard, thanks in advance.
CA_diffusion:
--config-name="RFdiffusion_CA_inference"
inference.model_runner="NRBStyleSelfCond"
inference.num_designs=2
inference.ckpt_path=${ckpt_path}
inference.output_prefix=${output_pref}
inference.design_startnum=1
+inference.two_template=true
+inference.three_template=true
inference.cautious=true
+inference.motif_only_2d=true
+inference.supply_motif_seq=true
inference.ligand="TEG"
inference.ij_visible="abc"
inference.input_pdb=${input_pdb}
inference.recenter_xt=true
preprocess.eye_frames=true
contigmap.contigs='["18,A1-3,31,B4,45"]'
denoiser.noise_scale_frame=0.01
denoiser.noise_scale_ca=0.05
diffuser.T=15
diffuser.r3.noise_scale=0.05
Refinement:
--config-name="test_ca_rfd_refinement"
inference.model_runner="NRBStyleSelfCond"
inference.num_designs=1
inference.ckpt_path="${ckpt_path}"
inference.output_prefix="${output_dir}/refined_${i}"
inference.input_pdb="${pdb_file}"
+inference.input_trb="${trb_file}"
+inference.refine=true
inference.refine_recycles=20
+inference.refine_w_ligand=true
inference.ligand="TEG"
++inference.two_template=true
++inference.three_template=true
++inference.cautious=true
++inference.motif_only_2d=true
++inference.supply_motif_seq=true
diffuser.T=15
+diffuser.so3_type=random
+diffuser.eucl_type=gaussian \
My theozyme looks very disorganized after the CA_diffusion and refinement steps. The rigidity is not maintained at all and there seems to be overlap of atoms in some cases. I could provide the input script I am using in both cases. I have changed the values of inference.refine_recycles in the range of 4 to 20, diffuser.T from 50 to 15, denoiser.noise_scale_frame from 0.05 to 0.01. none of these measures seem to give any useful results. Please help in this regard, thanks in advance.
CA_diffusion:
--config-name="RFdiffusion_CA_inference"
inference.model_runner="NRBStyleSelfCond"
inference.num_designs=2
inference.ckpt_path=${ckpt_path}
inference.output_prefix=${output_pref}
inference.design_startnum=1
+inference.two_template=true
+inference.three_template=true
inference.cautious=true
+inference.motif_only_2d=true
+inference.supply_motif_seq=true
inference.ligand="TEG"
inference.ij_visible="abc"
inference.input_pdb=${input_pdb}
inference.recenter_xt=true
preprocess.eye_frames=true
contigmap.contigs='["18,A1-3,31,B4,45"]'
denoiser.noise_scale_frame=0.01
denoiser.noise_scale_ca=0.05
diffuser.T=15
diffuser.r3.noise_scale=0.05
Refinement:
--config-name="test_ca_rfd_refinement"
inference.model_runner="NRBStyleSelfCond"
inference.num_designs=1
inference.ckpt_path="${ckpt_path}"
inference.output_prefix="${output_dir}/refined_${i}"
inference.input_pdb="${pdb_file}"
+inference.input_trb="${trb_file}"
+inference.refine=true
inference.refine_recycles=20
+inference.refine_w_ligand=true
inference.ligand="TEG"
++inference.two_template=true
++inference.three_template=true
++inference.cautious=true
++inference.motif_only_2d=true
++inference.supply_motif_seq=true
diffuser.T=15
+diffuser.so3_type=random
+diffuser.eucl_type=gaussian \