rc-foundry 0.1.7__tar.gz → 0.1.9__tar.gz

{rc_foundry-0.1.7 → rc_foundry-0.1.9}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: rc-foundry
-Version: 0.1.7
+Version: 0.1.9
 Summary: Shared utilities and training infrastructure for biomolecular structure prediction models.
 Author-email: Institute for Protein Design <contact@ipd.uw.edu>
 License: BSD 3-Clause License
@@ -97,11 +97,15 @@ Foundry provides tooling and infrastructure for using and training all classes o
 
 All models within Foundry rely on [AtomWorks](https://github.com/RosettaCommons/atomworks) - a unified framework for manipulating and processing biomolecular structures - for both training and inference. 
 
+
+> [!NOTE]
+> We have a slack now! Join for updates and to get your questions answered [here](https://join.slack.com/t/proteinmodelfoundry/shared_invite/zt-3kpwru8c6-nrmTW6LNHnSE7h16GNnfLA).
+
 ## Getting Started
 ### Quickstart guide
 **Installation**
 ```bash
-pip install rc-foundry[all]
+pip install "rc-foundry[all]"
 ```
 
 **Downloading weights** Models can be downloaded to a target folder with:

{rc_foundry-0.1.7 → rc_foundry-0.1.9}/README.md

@@ -4,11 +4,15 @@ Foundry provides tooling and infrastructure for using and training all classes o
 
 All models within Foundry rely on [AtomWorks](https://github.com/RosettaCommons/atomworks) - a unified framework for manipulating and processing biomolecular structures - for both training and inference. 
 
+
+> [!NOTE]
+> We have a slack now! Join for updates and to get your questions answered [here](https://join.slack.com/t/proteinmodelfoundry/shared_invite/zt-3kpwru8c6-nrmTW6LNHnSE7h16GNnfLA).
+
 ## Getting Started
 ### Quickstart guide
 **Installation**
 ```bash
-pip install rc-foundry[all]
+pip install "rc-foundry[all]"
 ```
 
 **Downloading weights** Models can be downloaded to a target folder with:

{rc_foundry-0.1.7 → rc_foundry-0.1.9}/examples/all.ipynb

@@ -91,6 +91,7 @@
     "config = RFD3InferenceConfig(\n",
     "    specification={\n",
     "        'length': 80,  # Generate 80-residue proteins\n",
+    "        'extra': {},  # We are not using any extra specifications here.\n",
     "    },\n",
     "    diffusion_batch_size=2,  # Generate 2 structures per batch\n",
     ")\n",

{rc_foundry-0.1.7 → rc_foundry-0.1.9}/models/rf3/src/rf3/cli.py

@@ -23,10 +23,19 @@ def fold(
         configure_minimal_inference_logging()
 
     # Find the RF3 configs directory relative to this file
-    # This file is at: models/rf3/src/rf3/cli.py
-    # Configs are at: models/rf3/configs/
-    rf3_package_dir = Path(__file__).parent.parent.parent  # Go up to models/rf3/
-    config_path = str(rf3_package_dir / "configs")
+    # In development: models/rf3/src/rf3/cli.py -> models/rf3/configs/
+    # When installed: site-packages/rf3/cli.py -> site-packages/rf3/configs/
+    rf3_file_dir = Path(__file__).parent
+
+    # Check if we're in installed mode (configs are sibling to this file)
+    # or development mode (configs are ../../../configs)
+    if (rf3_file_dir / "configs").exists():
+        # Installed mode
+        config_path = str(rf3_file_dir / "configs")
+    else:
+        # Development mode
+        rf3_package_dir = rf3_file_dir.parent.parent  # Go up to models/rf3/
+        config_path = str(rf3_package_dir / "configs")
 
     # Get all arguments
     args = ctx.params.get("args", []) + ctx.args

{rc_foundry-0.1.7 → rc_foundry-0.1.9}/models/rf3/src/rf3/inference.py

@@ -16,7 +16,9 @@ rootutils.setup_root(__file__, indicator=".project-root", pythonpath=True)
 
 load_dotenv(override=True)
 
-_config_path = os.path.join(os.environ["PROJECT_ROOT"], "models/rf3/configs")
+_config_path = os.path.join(
+    os.path.dirname(os.path.dirname(os.path.dirname(__file__))), "configs"
+)
 
 
 @hydra.main(

{rc_foundry-0.1.7 → rc_foundry-0.1.9}/models/rfd3/docs/demo.json

@@ -29,6 +29,7 @@
         "input": "./input_pdbs/1bna.pdb",
         "contig": "A1-10,/0,B15-24,/0,120-130",
         "length": "140-150",
-        "ori_token": [24,20,10]
+        "ori_token": [24,20,10],
+        "is_non_loopy": true
     }
 }

{rc_foundry-0.1.7 → rc_foundry-0.1.9}/models/rfd3/docs/na_binder_design.json

@@ -3,25 +3,29 @@
         "input": "./input_pdbs/1bna.pdb",
         "contig": "A1-10,/0,B15-24,/0,120-130",
         "length": "140-150",
-        "ori_token": [24,20,10]
+        "ori_token": [24,20,10],
+        "is_non_loopy": true
     },
     "ssDNA_basic": {
         "input": "./input_pdbs/5o4d.pdb",
         "contig": "A1-23,/0,120-130",
         "length": "143-153",
-        "ori_token": [-5,-10,8]
+        "ori_token": [-5,-10,8],
+        "is_non_loopy": true
     },
     "ssDNA_diffused_from_dsDNA_pdb":{
         "input": "./input_pdbs/1bna.pdb",
         "contig": "A1-10,/0,120-130",
         "length": "130-140",
-        "select_fixed_atoms": {"A1-10":""}
+        "select_fixed_atoms": {"A1-10":""},
+        "is_non_loopy": true
     },
     "RNA_basic": {
         "input": "./input_pdbs/1q75.pdb",
         "contig": "A1-15,/0,120-130",
         "length": "135-145",
-        "ori_token": [15,2,-4]
+        "ori_token": [15,2,-4],
+        "is_non_loopy": true
     },
     "dsDNA_complex": {
         "input": "./input_pdbs/2r5z.pdb",
@@ -36,7 +40,8 @@
         },
         "ori_token":[25,35,20],
         "select_hbond_acceptor": {"C16":"N7,O6", "D31-32":"N7", "D28-30":"OP1,OP2,O3',O5'"},
-        "select_hbond_donor": {"D31-32":"N6"}
+        "select_hbond_donor": {"D31-32":"N6"},
+        "is_non_loopy": true
 
     }
 }

{rc_foundry-0.1.7 → rc_foundry-0.1.9}/models/rfd3/docs/na_binder_design.md

@@ -30,7 +30,8 @@ The length attribute should be the sum of all polymer lengths. in this case (120
         "input": "./input_pdbs/1bna.pdb",
         "contig": "A1-10,/0,B15-24,/0,120-130",
         "length": "140-150",
-        "ori_token": [24,20,10]
+        "ori_token": [24,20,10],
+        "is_non_loopy": true
     }
 }
 ```
@@ -45,7 +46,8 @@ Similar to the previous example, but done for a PDB containing one DNA strand (A
         "input": "./input_pdbs/5o4d.pdb",
         "contig": "A1-23,/0,120-130",
         "length": "143-153",
-        "ori_token": [-5,-10,8]
+        "ori_token": [-5,-10,8],
+        "is_non_loopy": true
     }
 }
 ```
@@ -59,7 +61,8 @@ Similar to the previous example but the input PDB has a dsDNA. One of the chains
         "input": "./input_pdbs/1bna.pdb",
         "contig": "A1-10,/0,120-130",
         "length": "130-140",
-        "select_fixed_atoms": {"A1-10":""}
+        "select_fixed_atoms": {"A1-10":""},
+        "is_non_loopy": true
     }
 }
 ```
@@ -74,7 +77,8 @@ Example on RNA. Similar to the ssDNA example, example 2.
         "input": "./input_pdbs/1q75.pdb",
         "contig": "A1-15,/0,120-130",
         "length": "135-145",
-        "ori_token": [15,2,-4]
+        "ori_token": [15,2,-4],
+        "is_non_loopy": true
     }   
 }
 ```
@@ -104,7 +108,8 @@ To run this without warnings, you will need to install [hbplus](https://www.ebi.
         },
         "ori_token":[25,35,20],
         "select_hbond_acceptor": {"C16":"N7,O6", "D31-32":"N7", "D28-30":"OP1,OP2,O3',O5'"},
-        "select_hbond_donor": {"D31-32":"N6"}
+        "select_hbond_donor": {"D31-32":"N6"},
+        "is_non_loopy": true
 
     }
 }

{rc_foundry-0.1.7 → rc_foundry-0.1.9}/models/rfd3/docs/protein_binder_design.json

@@ -8,7 +8,8 @@
             "E64": "CD2,CZ",
             "E88": "CG,CZ",
             "E96": "CD1,CZ"
-        }
+       },
+       "is_non_loopy": true
     },
     "pdl1": {
         "dialect": 2,
@@ -19,6 +20,7 @@
             "A56": "CG,OH",
             "A115": "CG,SD",
             "A123": "CD2,OH"
-       }
+       },
+       "is_non_loopy": true
     }
 }

{rc_foundry-0.1.7 → rc_foundry-0.1.9}/models/rfd3/docs/protein_binder_design.md

@@ -5,6 +5,13 @@ RFD3 is a highly proficient protein binder designer. The following arguments hav
 - infer_ori_strategy: how RFD3 decides to place the origin of the generated protein binder with respect to the target. We find that using the "hotspots" strategy works best
 - select_hotspots: which atoms on the target should be bound (dictionary of residues on the target and atoms in those residues)
 
+In addition, we strongly recommend the following setting, which encourages the model to make more structured designs:
+- is_non_loopy: true
+
+Some useful command-line overrides to experiment with include `inference_sampler.step_scale` (defaults to 1.5) and
+`inference_sampler.gamma_0` (defaults to 0.6). Increasing the `step_scale` and decreasing `gamma_0` yields lower-temperature
+designs, which tends to increase designability and decrease diversity.
+
 If you would like to run the examples below, `protein_binder_design.json`, located in this directory, contains the example code. You can run it via:
 ```
 rfd3 design out_dir=inference_outputs/protein_binder/0 \
@@ -35,7 +42,8 @@ The input files for the different examples are provided in `foundry/models/rfd3/
             "E64": "CD2,CZ",
             "E88": "CG,CZ",
             "E96": "CD1,CZ",
-            }
+            },
+        "is_non_loopy": true
     },
     "pdl1": {
         "dialect": 2,
@@ -46,7 +54,8 @@ The input files for the different examples are provided in `foundry/models/rfd3/
             "A56": "CG,OH",
             "A115": "CG,SD",
             "A123": "CD2,OH",
-       }
+       },
+        "is_non_loopy": true
     }
 }
 ```

{rc_foundry-0.1.7 → rc_foundry-0.1.9}/models/rfd3/docs/symmetry.md

@@ -16,33 +16,40 @@ symmetry:
     is_unsym_motif: "Y1-11,Z16-25"
     is_symmetric_motif: true
 ```
-- `id`                : Symmetry group ID; e.g. "C3" for a cyclic protein with 3 subunits, "D2" for a dihedral protein with 2 subunits. Note that only C and D symmetry types are supported currently.
+- `id`                : Symmetry group ID. Supported symmetry types:
+  - **Cyclic (C)**: e.g. "C3" for a cyclic protein with 3 subunits
+  - **Dihedral (D)**: e.g. "D2" for a dihedral protein with 4 subunits (2×2)
+  - **Tetrahedral (T)**: "T" for tetrahedral symmetry with 12 subunits
+  - **Octahedral (O)**: "O" for octahedral symmetry with 24 subunits
+  - **Icosahedral (I)**: "I" for icosahedral symmetry with 60 subunits
 - `is_unsym_motif`    : Comma separated string list of contig/ligand names that should NOT be symmetrized (e.g. DNA strands). If not provided, all motifs are assumed to be symmetrized. See [Designs with motifs](#designs-with-motifs) section for details.
 - `is_symmetric_motif`: Boolean value whether the input motif is symmetric. Currently only symmetric input motifs are supported, therefore, `true` by default.
 
+> **⚠️ Memory Warning:** Memory requirements scale quadratically with the number of subunits. For larger complexes (especially T, O, and I symmetries), memory usage can become very high. Always use `diffusion_batch_size=1` for symmetry, and consider enabling `low_memory_mode=True` for higher-order symmetries. The memory footprint increases dramatically as both the number of subunits and the length of each subunit increase. Note: many higher order symmetires run out of memory even on large cards such as H200s, we are working on optimizations to make these networks more memory efficient. 
+
 
 ## Example command 
 You can run the following example command:
 ```
 ./src/modelhub/inference.py inference_sampler.kind=symmetry out_dir=logs/inference_outs/sym_demo/0 ckpt_path=$cur_ckpt inputs=./projects/aa_design/tests/test_data/sym_tests.json diffusion_batch_size=1 
 ```
-- `inference_sampler.kind`: Set `symmetry` to tern on symmetry mode.
-- `diffusion_batch_size`  : `8` by default, but it is recommended to set it to `1` for symmetry due to memory limitations.
-- `low_memory_mode`       : Additionally you can set this to `True` if you have memory constraints (e.g. "CUDA error: out of memory"). However, this will significantly slow the inference.
+- `inference_sampler.kind`: Set `symmetry` to turn on symmetry mode.
+- `diffusion_batch_size`  : **Must be set to `1` for all symmetry types** due to memory limitations. Memory scales quadratically with the number of subunits.
+- `low_memory_mode`       : **Strongly recommended** for T, O, and I symmetries. Set to `True` if you encounter memory errors (e.g. "CUDA error: out of memory"). Note that this will significantly slow the inference but is often necessary for larger complexes.
 
 
 ## Unconditional multimer design
 
-As mentioned above, we currently only support C and D symmetry types.
-The following provides a general overview of the types of symmetry and examples of how to run:
+The following provides a general overview of the supported symmetry types and examples of how to run:
 
-### Cyclic
-**Defaults:**
+### Cyclic (C)
+Cyclic symmetry with n-fold rotational symmetry around a single axis. Generates n identical subunits.
 
 ```json
 {
     "uncond_C15": {
         "length": 100,
+        "is_non_loopy": true,
         "symmetry": {
             "id": "C15"
         }
@@ -50,13 +57,14 @@ The following provides a general overview of the types of symmetry and examples
 }
 ```
 
-### Dihedrals
-**Defaults:**
+### Dihedral (D)
+Dihedral symmetry combines n-fold rotational symmetry with a 2-fold rotation perpendicular to the main axis. Generates 2n identical subunits.
 
 ```json
 {
     "uncond_D4": { 
         "length": 100,
+        "is_non_loopy": true,
         "symmetry": {
             "id": "D4"
         }
@@ -64,9 +72,57 @@ The following provides a general overview of the types of symmetry and examples
 }
 ```
 
+### Tetrahedral (T)
+Tetrahedral symmetry based on the symmetry of a tetrahedron. Generates 12 identical subunits.
+
+```json
+{
+    "uncond_T": { 
+        "length": 100,
+        "symmetry": {
+            "id": "T"
+        }
+    }
+}
+```
+
+### Octahedral (O)
+Octahedral symmetry based on the symmetry of a cube/octahedron. Generates 24 identical subunits.
+
+```json
+{
+    "uncond_O": { 
+        "length": 100,
+        "symmetry": {
+            "id": "O"
+        }
+    }
+}
+```
+
+### Icosahedral (I)
+Icosahedral symmetry based on the symmetry of an icosahedron/dodecahedron. Generates 60 identical subunits. This is the largest point group symmetry and is commonly found in viral capsids.
+
+```json
+{
+    "uncond_I": { 
+        "length": 100,
+        "symmetry": {
+            "id": "I"
+        }
+    }
+}
+```
+
+> **⚠️ Memory Warning:** Memory requirements increase dramatically with the number of subunits and subunit length. The memory footprint scales approximately quadratically with the number of subunits due to pairwise interactions. For higher-order symmetries (T: 12 subunits, O: 24 subunits, I: 60 subunits), it is **essential** to:
+> - Set `diffusion_batch_size=1` (required for all symmetry types)
+> - Enable `low_memory_mode=True` for T, O, and I symmetries
+> - Consider reducing subunit length if you encounter out-of-memory errors
+> - For icosahedral symmetry (60 subunits), expect memory usage to be 25-100× higher than a single chain design
+
 ## Designs with motifs
 
-As mentioned above, symmetry sampling currently only supports pre-symmetrized motifs around the origin. Therefore, `is_symmetric_motif` is set to `true` by default. 
+Symmetry sampling currently only supports pre-symmetrized motifs around the origin. Therefore, `is_symmetric_motif` is set to `true` by default. 
 The following are example JSON specifications for different symmetric motif scaffolding. You can also find the corresponding input PDBs in `docs/input_pdbs/symmetry_examples`. Although we only give JSON examples, you can also use YAML for everything shown below.   
 
 The tasks that these examples describe are as follows:
@@ -122,7 +178,7 @@ The tasks that these examples describe are as follows:
             "M52": "CG,SD,CE"
         }
     },
-    "uncond_unsym_C3_6t8h": {
+    "unsym_C3_6t8h": {
         "symmetry": {
             "id": "C3",
             "is_symmetric_motif": true,
@@ -130,7 +186,8 @@ The tasks that these examples describe are as follows:
         },
         "input": "symmetry_examples/6t8h_C3.pdb",
         "contig": "150-150,/0,Y1-11,/0,Z16-25",
-        "length": null
+        "length": null,
+        "is_non_loopy": true
     }
 }
 ```

{rc_foundry-0.1.7 → rc_foundry-0.1.9}/models/rfd3/src/rfd3/engine.py

@@ -21,11 +21,13 @@ from rfd3.constants import SAVED_CONDITIONING_ANNOTATIONS
 from rfd3.inference.datasets import (
     assemble_distributed_inference_loader_from_json,
 )
-from rfd3.inference.input_parsing import DesignInputSpecification
+from rfd3.inference.input_parsing import (
+    DesignInputSpecification,
+    ensure_input_is_abspath,
+)
 from rfd3.model.inference_sampler import SampleDiffusionConfig
 from rfd3.utils.inference import (
     ensure_inference_sampler_matches_design_spec,
-    ensure_input_is_abspath,
 )
 from rfd3.utils.io import (
     CIF_LIKE_EXTENSIONS,
@@ -391,7 +393,13 @@ class RFD3InferenceEngine(BaseInferenceEngine):
         design_specifications = {}
         for prefix, example_spec in inputs.items():
             # Record task name in the specification
-            example_spec["extra"]["task_name"] = prefix
+            if isinstance(example_spec, DesignInputSpecification):
+                example_spec.extra = example_spec.extra or {}
+                example_spec.extra["task_name"] = prefix
+            else:
+                if "extra" not in example_spec:
+                    example_spec["extra"] = {}
+                example_spec["extra"]["task_name"] = prefix
 
             # ... Create n_batches for example
             for batch_id in range((n_batches) if exists(n_batches) else 1):

{rc_foundry-0.1.7 → rc_foundry-0.1.9}/models/rfd3/src/rfd3/inference/datasets.py

@@ -14,8 +14,8 @@ from atomworks.ml.transforms.base import Compose, Transform
 from omegaconf import DictConfig, OmegaConf
 from rfd3.inference.input_parsing import (
     DesignInputSpecification,
+    ensure_input_is_abspath,
 )
-from rfd3.utils.inference import ensure_input_is_abspath
 from torch.utils.data import (
     DataLoader,
     SequentialSampler,

{rc_foundry-0.1.7 → rc_foundry-0.1.9}/models/rfd3/src/rfd3/inference/input_parsing.py

@@ -5,6 +5,7 @@ import os
 import time
 import warnings
 from contextlib import contextmanager
+from os import PathLike
 from typing import Any, Dict, List, Optional, Union
 
 import numpy as np
@@ -1121,3 +1122,33 @@ def accumulate_components(
     if atom_array_accum.bonds is None:
         atom_array_accum.bonds = BondList(atom_array_accum.array_length())
     return atom_array_accum
+
+
+def ensure_input_is_abspath(args: Dict[str, Any], path: PathLike | None):
+    """
+    Ensures the input source is an absolute path if exists, if not it will convert
+
+    args:
+        args: Inference specification for atom array
+        path: None or file to which the input is relative to.
+    """
+    if isinstance(args, str):
+        raise ValueError(
+            "Expected args to be a dictionary, got a string: {}. If you are using an input JSON ensure it contains dictionaries of arguments".format(
+                args
+            )
+        )
+    if "input" not in args or not exists(args["input"]):
+        return args
+    input = str(args["input"])
+    if not os.path.isabs(input):
+        if path is None:
+            raise ValueError(
+                "Input path is relative, but no base path was provided to resolve it against."
+            )
+        input = os.path.abspath(os.path.join(os.path.dirname(str(path)), input))
+        logger.info(
+            f"Input source path is relative, converted to absolute path: {input}"
+        )
+        args["input"] = input
+    return args

{rc_foundry-0.1.7 → rc_foundry-0.1.9}/models/rfd3/src/rfd3/inference/symmetry/atom_array.py

@@ -1,3 +1,5 @@
+import string
+
 import numpy as np
 from rfd3.inference.symmetry.frames import (
     decompose_symmetry_frame,
@@ -7,6 +9,68 @@ from rfd3.inference.symmetry.frames import (
 FIXED_TRANSFORM_ID = -1
 FIXED_ENTITY_ID = -1
 
+# Alphabet for chain ID generation (uppercase letters only, per wwPDB convention)
+_CHAIN_ALPHABET = string.ascii_uppercase
+
+
+def index_to_chain_id(index: int) -> str:
+    """
+    Convert a zero-based index to a chain ID following wwPDB convention.
+
+    The naming follows the wwPDB-assigned chain ID system:
+    - 0-25: A-Z (single letter)
+    - 26-701: AA-ZZ (double letter)
+    - 702-18277: AAA-ZZZ (triple letter)
+    - And so on...
+
+    This is similar to Excel column naming (A, B, ..., Z, AA, AB, ...).
+
+    Arguments:
+        index: zero-based index (0 -> 'A', 25 -> 'Z', 26 -> 'AA', etc.)
+    Returns:
+        chain_id: string chain identifier
+    """
+    if index < 0:
+        raise ValueError(f"Chain index must be non-negative, got {index}")
+
+    result = ""
+    remaining = index
+
+    # Convert to bijective base-26 (like Excel columns)
+    while True:
+        result = _CHAIN_ALPHABET[remaining % 26] + result
+        remaining = remaining // 26 - 1
+        if remaining < 0:
+            break
+
+    return result
+
+
+def chain_id_to_index(chain_id: str) -> int:
+    """
+    Convert a chain ID back to a zero-based index.
+
+    Inverse of index_to_chain_id.
+
+    Arguments:
+        chain_id: string chain identifier (e.g., 'A', 'Z', 'AA', 'AB')
+    Returns:
+        index: zero-based index
+    """
+    if not chain_id or not all(c in _CHAIN_ALPHABET for c in chain_id):
+        raise ValueError(f"Invalid chain ID: {chain_id}")
+
+    # Offset for all shorter chain IDs (26 + 26^2 + ... + 26^(len-1))
+    offset = sum(26**k for k in range(1, len(chain_id)))
+
+    # Value within the current length group (standard base-26)
+    value = 0
+    for char in chain_id:
+        value = value * 26 + _CHAIN_ALPHABET.index(char)
+
+    return offset + value
+
+
 ########################################################
 # Symmetry annotations
 ########################################################
@@ -247,11 +311,13 @@ def reset_chain_ids(atom_array, start_id):
     Reset the chain ids and pn_unit_iids of an atom array to start from the given id.
     Arguments:
         atom_array: atom array with chain_ids and pn_unit_iids annotated
+        start_id: starting chain ID (e.g., 'A')
     """
     chain_ids = np.unique(atom_array.chain_id)
-    new_chain_range = range(ord(start_id), ord(start_id) + len(chain_ids))
-    for new_id, old_id in zip(new_chain_range, chain_ids):
-        atom_array.chain_id[atom_array.chain_id == old_id] = chr(new_id)
+    start_index = chain_id_to_index(start_id)
+    for i, old_id in enumerate(chain_ids):
+        new_id = index_to_chain_id(start_index + i)
+        atom_array.chain_id[atom_array.chain_id == old_id] = new_id
     atom_array.pn_unit_iid = atom_array.chain_id
     return atom_array
 
@@ -259,15 +325,18 @@ def reset_chain_ids(atom_array, start_id):
 def reannotate_chain_ids(atom_array, offset, multiplier=0):
     """
     Reannotate the chain ids and pn_unit_iids of an atom array.
+
+    Uses wwPDB-style chain IDs (A-Z, AA-ZZ, AAA-ZZZ, ...) to support
+    any number of chains.
+
     Arguments:
         atom_array: protein atom array with chain_ids and pn_unit_iids annotated
-        offset: offset to add to the chain ids
-        multiplier: multiplier to add to the chain ids
+        offset: offset to add to the chain ids (typically num_chains in ASU)
+        multiplier: multiplier for the offset (typically transform index)
     """
-    chain_ids_int = (
-        np.array([ord(c) for c in atom_array.chain_id]) + offset * multiplier
-    )
-    chain_ids = np.array([chr(id) for id in chain_ids_int], dtype=str)
+    chain_ids_indices = np.array([chain_id_to_index(c) for c in atom_array.chain_id])
+    new_indices = chain_ids_indices + offset * multiplier
+    chain_ids = np.array([index_to_chain_id(idx) for idx in new_indices], dtype="U4")
     atom_array.chain_id = chain_ids
     atom_array.pn_unit_iid = chain_ids
     return atom_array

{rc_foundry-0.1.7 → rc_foundry-0.1.9}/models/rfd3/src/rfd3/inference/symmetry/checks.py

@@ -24,7 +24,16 @@ def check_symmetry_config(
     assert sym_conf.id, "symmetry_id is required. e.g. {'id': 'C2'}"
     # if unsym motif is provided, check that each motif name is in the atom array
 
+    is_motif_atom = get_motif_features(atom_array)["is_motif_atom"]
     is_unsym_motif = np.zeros(atom_array.shape[0], dtype=bool)
+
+    if not is_motif_atom.any():
+        sym_conf.is_symmetric_motif = None
+        ranked_logger.warning(
+            "No motifs found in atom array. Setting is_symmetric_motif to None."
+        )
+        return sym_conf
+
     if sym_conf.is_unsym_motif:
         assert (
             src_atom_array is not None
@@ -36,21 +45,20 @@ def check_symmetry_config(
             if (sm and n not in sm.split(",")) and (n not in atom_array.src_component):
                 raise ValueError(f"Unsym motif {n} not found in atom_array")
 
-    is_motif_token = get_motif_features(atom_array)["is_motif_token"]
     if (
-        is_motif_token[~is_unsym_motif].any()
+        is_motif_atom[~is_unsym_motif].any()
         and not sym_conf.is_symmetric_motif
         and not has_dist_cond
     ):
         raise ValueError(
-            "Asymmetric motif inputs should be distance constrained."
-            "Use atomwise_fixed_dist to constrain the distance between the motif atoms."
+            "Asymmetric motif inputs are not supported yet. Please provide a symmetric motif."
         )
 
     if partial and not sym_conf.is_symmetric_motif:
         raise ValueError(
             "Partial diffusion with symmetry is only supported for symmetric inputs."
         )
+    return sym_conf
 
 
 def check_atom_array_is_symmetric(atom_array):