EModelX

EModelX is a method for automatic cryo-EM protein complex structure modeling.

Colab

EModelX can be run in Colab: Minimal Example

Environment

conda env create -f EModelX.yml
For EModelX(+AF), you may need to run AlphaFold following https://github.com/deepmind/alphafold or get AlphaFold-predicted single-chain structures from AlphaFoldDB (https://alphafold.ebi.ac.uk/).

Download NN model weights

Download NN model weights from https://drive.google.com/file/d/13BKzEBfL0uubYgcJJTGSZ-PkO9hXAB9X/view?usp=drive_link and place them in ./models.

If you don't have standard fasta file

Run:
python get_fasta.py gen_fasta_path 'fasta_name_1;chain_num_1;sequence_1' 'fasta_name_2;chain_num_2;sequence_2' ...;...;... 'fasta_name_n;chain_num_n;sequence_n'
, where gen_fasta_path is the path to save the generated fasta file (e.g. data/inputs/fastas/7w72)
, chain_num_n should be an integer that indicates how many chains belong to this sequence
, and sequence_n should be a protein sequence
, and quotation marks ' or " would be ignored
, and make sure ;s are only used as separtion marks.

Minimal Example: Protein complex structure modeling for a cryo-EM map

For EModelX:
python run.py --protocol=temp_free --EM_map=data/inputs/maps/emd_32336.map.gz --fasta=data/inputs/fastas/7w72 --output_dir=data/outputs --run_pulchra --pulchra_path modules/pulchra304/src/pulchra

For EModelX(+AF), now we support automatically download the most similar alphafold structure from alphafoldDB within a certain similarity cutoff:
python run.py --protocol=temp_flex --EM_map=data/inputs/maps/emd_32336.map.gz --fasta=data/inputs/fastas/7w72 --template_dir=data/inputs/templates --download_afdb --afdb_allow_seq_id 0.95 --output_dir=data/outputs --run_pulchra --pulchra_path=modules/pulchra304/src/pulchra
, where you can replace --EM_map with your target EM map
, and --fasta with your target fasta
, and --template_dir: directory of the template folder, only needed when --protocol == temp_flex
, and --output_dir: the output directory for modeling results

Notice: If you want to run EModelX(+AF), please either use --download_afdb or prepare the --template_dir by yourself. To prepare the --template_dir by yourself, please follow our example in ./inputs/templates. More detailedly, you need to prepare a directory with a consistent name matching the header lines in the fasta file, and the PDB file needed to be renamed as ranked_0.pdb.

To enable postprocessing by phenix.real_space_refinement

Environment

Install phenix following https://phenix-online.org/ into a directory, e.g. modules/phenix-1.20.1-4487

Example

For EModelX:
python run.py --protocol=temp_free --EM_map=data/inputs/maps/emd_32336.map.gz --fasta=data/inputs/fastas/7w72 --output_dir=data/outputs --run_pulchra --pulchra_path=modules/pulchra304/src/pulchra --run_phenix --phenix_act=modules/phenix-1.20.1-4487/phenix_env.sh --resolution=3.1
For EModelX(+AF):
python run.py --protocol=temp_flex --EM_map=data/inputs/maps/emd_32336.map.gz --fasta=data/inputs/fastas/7w72 --template_dir=data/inputs/templates --output_dir=data/outputs --download_afdb --afdb_allow_seq_id 0.95 --run_pulchra --pulchra_path=modules/pulchra304/src/pulchra --run_phenix --phenix_act=modules/phenix-1.20.1-4487/phenix_env.sh --resolution=3.1

Web Server

EModelX's web server is accessible at https://bio-web1.nscc-gz.cn/app/EModelX