Openfold addition #49

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged

hlasimpk merged 12 commits into main from openfold_addition

Feb 4, 2026

README.md

            
                      Original file line number
                      Diff line number
                      Diff line change
                  
    @@ -4,7 +4,7 @@
  
    ![Coverage](https://raw.githubusercontent.com/rigdenlab/ABCFold/refs/heads/main/.blob/coverage.svg)

    Scripts to run AlphaFold3, Boltz, Chai-1 and Protenix with MMseqs2 Multiple sequence alignments (MSAs) and custom templates.

    Scripts to run AlphaFold3, Boltz, Chai-1, OpenFold3 and Protenix with MMseqs2 Multiple sequence alignments (MSAs) and custom templates.

    ## Table of Contents

    - [Installation](#installation)

    @@ -26,7 +26,7 @@ micromamba activate abcfold
  
    > **Environment note**

    >

    > - Your main Python environment can be Conda, micromamba, or virtualenv.

    > - `abcfold` will automatically create **internal micromamba environments** to run Boltz, Chai-1 and Protenix safely, therefore a micromamba installation is required.

    > - `abcfold` will automatically create **internal micromamba environments** to run Boltz, Chai-1, OpenFold3 and Protenix safely, therefore a micromamba installation is required.

    > - This prevents package conflicts with your main environment and ensures reproducible results.

    @@ -57,7 +57,7 @@ python -m pre_commit install
  
    ### Running ABCfold

    ABCFold will run Alphafold3, Boltz, Chai-1 and Protenix consecutively. The program takes an input of a JSON in the Alphafold3 format (For full instruction on how to format this, click [here](https://github.com/google-deepmind/alphafold3/blob/main/docs/input.md)). An example JSON is shown below:

    ABCFold will run Alphafold3, Boltz, Chai-1, OpenFold3 and Protenix consecutively. The program takes an input of a JSON in the Alphafold3 format (For full instruction on how to format this, click [here](https://github.com/google-deepmind/alphafold3/blob/main/docs/input.md)). An example JSON is shown below:

    ```json

    {

    @@ -76,11 +76,11 @@ ABCFold will run Alphafold3, Boltz, Chai-1 and Protenix consecutively. The progr
  
    }

    ```

    Please make sure you have AlphaFold3 installed on your system (Instructions [here](https://github.com/google-deepmind/alphafold3/blob/main/docs/installation.md)) and have procured the model parameters. Boltz and Chai-1 are installed upon runtime.

    Please make sure you have AlphaFold3 installed on your system (Instructions [here](https://github.com/google-deepmind/alphafold3/blob/main/docs/installation.md)) and have procured the model parameters. Boltz, Chai-1, OpenFold3 and Protenix are installed upon runtime.

    For the majority of jobs, ABCFold can be run as follows:

    ```bash

    abcfold <input_json>  <output_dir> -abcp --mmseqs2 --model_params <path_to_af3_model_params>

    abcfold <input_json>  <output_dir> -abcop --mmseqs2 --model_params <path_to_af3_model_params>

    ```

    > [!NOTE]

    > `--model_params` is stored after the first run, therefore subsequent ABCFold jobs don't require this flag.

    @@ -100,7 +100,7 @@ However, there you may wish to use the following flags to add run time options s
  
    #### Main arguments

    - `<input_json>`: Path to the input AlphaFold3 JSON file.

    - `<output_dir>`: Path to the output directory.

    - `-a`, `-b`, `-c`, `-p` (`--alphafold3`, `--boltz`,`--chai1`,`--protenix`): Flags to run Alphafold3, Boltz, Chai-1 and Protenix respectively. If none of these flags are provided, Alphafold3 will be run by default.

    - `-a`, `-b`, `-c`, `-o`, `-p` (`--alphafold3`, `--boltz`,`--chai1`, `--openfold3`, `--protenix`): Flags to run Alphafold3, Boltz, Chai-1, OpenFold3 and Protenix respectively. If none of these flags are provided, Alphafold3 will be run by default.

    - `--mmseqs2`: [optional] Flag to use MMseqs2 MSAs and templates (if specified).

    - `--mmseqs_database`: [optional] The path to the database used by a local copy of MMSeqs2, provided mmseqs is installed, the inclusion of this flag allows MMseqs2 to be run locally.

    - `--override`: [optional] Flag to override the existing output directory.

    @@ -114,6 +114,9 @@ However, there you may wish to use the following flags to add run time options s
  
    - `--af3_sif_path`: [optional] Path to sif file if using an AlphaFold3 singularity instead of Docker

    - `--use_af3_template_search`[optional] If providing your own custom MSA or you've ran `--mmseqs2`, allow Alphafold3 to search for templates

    #### OpenFold3 arguments

    - `--inference_ckpt_path` [optional] Path for model checkpoint to be used for inference. If not specified, will attempt to find or download parameters in ~/.openfold3/

    #### Template arguments

    - `--templates`: Flag to enable a template search

    @@ -132,19 +135,19 @@ However, there you may wish to use the following flags to add run time options s
  
    If you wanted to provide a custom template, `custom_a.pdb` for your protein sequence with the ID `A` and you have your template has two chains: chain `A` and chain `B` and chain `B` is what you want the template to be, you could run:

    ```bash

    abcfold <input_json>  <output_dir> -abcp --mmseqs2 --custom_template custom_a.pdb  --custom_template_chain B --target_id A

    abcfold <input_json>  <output_dir> -abcop --mmseqs2 --custom_template custom_a.pdb  --custom_template_chain B --target_id A

    ```

    If you had multiple IDs in your input sequence, multiple template files and you wanted to provide 3 custom templates, chain `A` from `custom_a.pdb`, chain `B` from `custom_b.pdb`, and chain B from `custom_c.pdb`, where `custom_a.pdb` and `custom_b.pdb` correspond to the ID `A` and `custom_c.pdb` corresponds to the ID `B`, you could run:

    ```bash

    abcfold <input_json>  <output_dir> -abcp --mmseqs2 --custom_template custom_a.pdb custom_b.pdb custom_c.pdb --custom_template_chain A B B --target_id A A B

    abcfold <input_json>  <output_dir> -abcop --mmseqs2 --custom_template custom_a.pdb custom_b.pdb custom_c.pdb --custom_template_chain A B B --target_id A A B

    ```

    ### Output

    ABCFold will output the AlphaFold, Boltz and/or Chai models in the `<output_dir>`, it will also produce an output page containing a results table and informative [PAE viewer](https://gitlab.gwdg.de/general-microbiology/pae-viewer). This is opened automatically in your default browser unless the `--no_server` or `--no_visuals` flags are used.

    ABCFold will output the AlphaFold, Boltz, Chai, OpenFold3 and Protenix models in the `<output_dir>`, it will also produce an output page containing a results table and informative [PAE viewer](https://gitlab.gwdg.de/general-microbiology/pae-viewer). This is opened automatically in your default browser unless the `--no_server` or `--no_visuals` flags are used.

    Unless the `--no_visuals` flag is used, you can then open the output pages by running:

    @@ -168,7 +171,7 @@ you will find `open_output.py` in your `<output_dir>`. This needs to be run from
  
    Below are scripts for adding MMseqs2 MSAs and custom templates to AlphaFold3 input JSON files.

    > [!WARNING]

    > These scripts will only modify the input JSON files, I.E. they will NOT run AlphaFold3, Boltz, Chai-1 and Protenix.

    > These scripts will only modify the input JSON files, I.E. they will NOT run AlphaFold3, Boltz, Chai-1, OpenFold3 and Protenix.

    ### Adding MMseqs2 MSAs and templates

    @@ -321,7 +324,12 @@ If the identical sequences are given as seperate entities (as shown below) you w
  
    }

    ```

    Additionally, Boltz currently lacks the ability to create linked-ligands and therefore covalent bonds between the chain/ligand will be missing.

    Additionally, Boltz currently lack the ability to create linked-ligands and therefore covalent bonds between the chain/ligand will be missing.

    #### OpenFold3 limitations

    Currently OpenFold3 is unable to create bonded-pairs or alter the number of recycles. Therefore these options will be ignored if provided.

    ## Contributing

abcfold/abcfold.py

            
                      Original file line number
                      Diff line number
                      Diff line change
                  
    @@ -14,6 +14,7 @@
  
                                        boltz_argparse_util, chai_argparse_util,

                                        custom_template_argpase_util,

                                        main_argpase_util, mmseqs2_argparse_util,

                                        openfold_argparse_util,

                                        prediction_argparse_util,

                                        protenix_argparse_util,

                                        raise_argument_errors,

    @@ -27,6 +28,7 @@
  
    from abcfold.output.boltz import BoltzOutput

    from abcfold.output.chai import ChaiOutput

    from abcfold.output.file_handlers import superpose_models

    from abcfold.output.openfold3 import OpenfoldOutput

    from abcfold.output.protenix import ProtenixOutput

    from abcfold.output.utils import (get_gap_indicies, insert_none_by_minus_one,

                                      make_dummy_m8_file)

    @@ -192,7 +194,7 @@ def run(args, config, defaults, config_file):
  
                template_hits_path = None

                if args.templates and args.mmseqs2:

                    template_hits_path = temp_dir.joinpath("all_chain.m8")

                    template_hits_path = temp_dir.joinpath("all_chains.m8")

                elif args.templates:

                    template_hits_path = make_dummy_m8_file(run_json, temp_dir)

    @@ -230,6 +232,30 @@ def run(args, config, defaults, config_file):
  
                    outputs.append(po)

                successful_runs.append(protenix_success)

            if args.openfold3:

                from abcfold.openfold3.run_openfold3 import run_openfold

                template_hits_path = None

                if args.templates and args.mmseqs2:

                    template_hits_path = temp_dir.joinpath("all_chains.m8")

                openfold_success = run_openfold(

                    input_json=run_json,

                    output_dir=args.output_dir,

                    save_input=args.save_input,

                    number_of_models=args.number_of_models,

                    template_hits_path=template_hits_path,

                    input_ckpt=args.inference_ckpt_path

                )

                if openfold_success:

                    openfold_output_dirs = list(args.output_dir.glob("openfold_results*"))

                    oo = OpenfoldOutput(

                        openfold_output_dirs, input_params, name, args.save_input

                    )

                    outputs.append(oo)

                successful_runs.append(openfold_success)

            if args.no_visuals:

                logger.info("Visuals disabled")

                return

    @@ -333,6 +359,34 @@ def run(args, config, defaults, config_file):
  
                                )

                                chai_models["models"].append(model_data)

            openfold_models = {"models": []}

            if args.openfold3:

                if openfold_success:

                    programs_run.append("OpenFold3")

                    for seed in oo.output.keys():

                        for idx in oo.output[seed].keys():

                            if idx >= 0:

                                model = oo.output[seed][idx]["cif"]

                                model.check_clashes()

                                score_file = oo.output[seed][idx]["scores"]

                                plddt = model.residue_plddts

                                pae = oo.output[seed][idx]["af3_pae"]

                                if len(indicies) > 0:

                                    plddt = insert_none_by_minus_one(

                                        indicies[index_counter], plddt

                                    )

                                index_counter += 1

                                model_data = get_model_data(

                                    model,

                                    plot_dict,

                                    "OpenFold3",

                                    plddt,

                                    pae,

                                    score_file,

                                    args.output_dir,

                                )

                                openfold_models["models"].append(model_data)

            protenix_models = {"models": []}

            if args.protenix:

                if protenix_success:

    @@ -365,6 +419,7 @@ def run(args, config, defaults, config_file):
  
                alphafold_models["models"] +

                boltz_models["models"] +

                chai_models["models"] +

                openfold_models["models"] +

                protenix_models["models"]

            )

    @@ -379,6 +434,8 @@ def run(args, config, defaults, config_file):
  
                    output_name = "boltz_model_" + model["model_id"][-1] + ".cif"

                elif model["model_source"] == "Chai-1":

                    output_name = "chai_model_" + model["model_id"][-1] + ".cif"

                elif model["model_source"] == "OpenFold3":

                    output_name = "openfold_model_" + model["model_id"][-1] + ".cif"

                elif model["model_source"] == "Protenix":

                    output_name = "protenix_model_" + model["model_id"][-1] + ".cif"

                shutil.copy(

    @@ -471,12 +528,12 @@ def run(args, config, defaults, config_file):
  
    def main():

        """

        Run AlphaFold3 / Boltz / Chai-1 / Protenix

        Run AlphaFold3 / Boltz / Chai-1 / OpenFold3 / Protenix

        """

        import argparse

        parser = argparse.ArgumentParser(

            description="Run AlphaFold3 / Boltz / Chai-1 / Protenix"

            description="Run AlphaFold3 / Boltz / Chai-1 / OpenFold3 / Protenix"

        )

        defaults = {}

    @@ -492,6 +549,7 @@ def main():
  
        parser = alphafold_argparse_util(parser)

        parser = boltz_argparse_util(parser)

        parser = chai_argparse_util(parser)

        parser = openfold_argparse_util(parser)

        parser = protenix_argparse_util(parser)

        parser = mmseqs2_argparse_util(parser)

        parser = custom_template_argpase_util(parser)

abcfold/alphafold3/check_install.py

-Original file line number
+Diff line change
@@ Expand Up / @@ -63,7 +63,7 @@ def generate_test_command(interactive: bool = True, @@
         Returns:
             str: The Alphafold3 help command
         """
-        if sif_path:
+        if sif_path is not None and sif_path != "None":
             return f"""
         singularity exec \
         {sif_path} \
@@ Expand All @@
         """
         Generate the Alphafold3 version command
         """
-        if sif_path:
+        if sif_path is not None and sif_path != "None":
             return f"""
         singularity exec \
         {sif_path} \
@@ Expand Down @@

abcfold/alphafold3/run_alphafold3.py

-Original file line number
+Diff line change
@@ Expand Up / @@ -104,7 +104,7 @@ def generate_af3_cmd( @@
         input_json = Path(input_json)
         output_dir = Path(output_dir)
-        if sif_path is not None:
+        if sif_path is not None and sif_path != "None":
             return f"""
             singularity exec \
             --nv \
@@ Expand Down @@

abcfold/argparse_utils.py

            
                      Original file line number
                      Diff line number
                      Diff line change
                  
    @@ -150,6 +150,22 @@ def protenix_argparse_util(parser):
  
        return parser

    def openfold_argparse_util(parser):

        parser.add_argument(

            "-o",

            "--openfold3",

            action="store_true",

            help="Run OpenFold 3",

        )

        parser.add_argument(

            "--inference_ckpt_path",

            help=dedent("Path for model checkpoint to be used for inference. \

        If not specified, will attempt to find or download parameters to \

         ~/.openfold3/")

        )

        return parser

    def alphafold_argparse_util(parser):

        parser.add_argument(

            "--database",

    @@ -213,9 +229,15 @@ def visuals_argparse_util(parser):
  
    def raise_argument_errors(args):

        if not args.alphafold3 and not args.boltz and not args.chai1 and not args.protenix:

        if (

            not args.alphafold3

            and not args.boltz

            and not args.chai1

            and not args.protenix

            and not args.openfold3

        ):

            logger.info(

                dedent("Neither AlphaFold3, Boltz, Chai-1, or Protenix selected. \

                dedent("None of AlphaFold3, Boltz, Chai-1, Protenix or OpenFold3 selected. \

                Running AlphaFold3 by default")

            )

            args.alphafold3 = True

abcfold/backend_envs.py

-Original file line number
+Diff line change
@@ Expand Up / @@ -99,9 +99,31 @@ def pip_install(self, packages: list[str]): @@
             self._run([
                 "run",
                 "-n", self.env_name,
-                "pip", "install", *packages,
+                "python", "-m", "pip", "install",
+                "--break-system-packages",
+                *packages,
             ])
+        def conda_install(
+            self,
+            packages: list[str],
+            *,
+            channels: Optional[list[str]] = None,
+        ):
+            cmd = [
+                "install",
+                "-y",
+                "-n", self.env_name,
+            ]
+            if channels:
+                for ch in channels:
+                    cmd.extend(["-c", ch])
+            cmd.extend(packages)
+            self._run(cmd)
         def run(self,
                 command: list[str],
                 capture_output: bool = False,
@@ Expand Down @@

abcfold/boltz/check_install.py

-Original file line number
+Diff line change
@@ Expand Up / @@ -14,7 +14,7 @@ def ensure_boltz_env(): @@
         # 1. Ensure env exists
         env.create(python_version="3.11")
-        # 2. Check installed chai version
+        # 2. Check installed boltz version
         installed = env.get_installed_version("boltz")
         if installed != BOLTZ_VERSION:
@@ Expand Down @@

abcfold/html/abcfold_vue.js

-Original file line number
+Diff line change
@@ Expand Up / @@ -60,6 +60,8 @@ Vue.component('abc-table', { @@
                         return 'btn-source3';
                     case 'Protenix':
                         return 'btn-source4';
+                    case 'OpenFold3':
+                        return 'btn-source5';
                     default:
                         return 'btn-default';
                 }
@@ Expand Down @@

abcfold/html/html_utils.py

-Original file line number
+Diff line change
@@ Expand Up / @@ -14,6 +14,7 @@ @@
     from abcfold.output.boltz import BoltzOutput
     from abcfold.output.chai import ChaiOutput
     from abcfold.output.file_handlers import ConfidenceJsonFile, NpzFile
+    from abcfold.output.openfold3 import OpenfoldOutput
     from abcfold.output.protenix import ProtenixOutput
     from abcfold.plots.pae_plot import create_pae_plots
     from abcfold.plots.plddt_plot import plot_plddt
@@ Expand Down Expand Up / @@ -278,6 +279,11 @@ def get_all_cif_files(outputs) -> Dict[str, list]: @@
                     if "Chai-1" not in method_cif_objs:
                         method_cif_objs["Chai-1"] = []
                     method_cif_objs["Chai-1"].extend(output.cif_files[seed])
+            elif isinstance(output, OpenfoldOutput):
+                for seed in output.seeds:
+                    if "Openfold3" not in method_cif_objs:
+                        method_cif_objs["Openfold3"] = []
+                    method_cif_objs["Openfold3"].extend(output.cif_files[seed])
             elif isinstance(output, ProtenixOutput):
                 for seed in output.seeds:
                     if "Protenix" not in method_cif_objs:
@@ Expand Down @@

abcfold/html/style.css

-Original file line number
+Diff line change
@@ Expand Up / @@ -486,6 +486,16 @@ rect{ @@
         transition: background-color 0.3s ease;
     }
+    .btn-source5 {
+        background-color: #E89EB8;
+        color: #3b3b3d;
+        border: none;
+        border-radius: 12px;
+        padding: 7px 15px;
+        cursor: pointer;
+        transition: background-color 0.3s ease;
+    }
     .btn-default {
         background-color: rgb(199, 195, 195);
         color: white;
@@ Expand Down @@

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Openfold addition #49

Uh oh!

Diff view

Diff view

There are no files selected for viewing

Uh oh!

Uh oh!