SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2409.17265 · ICML · 2024

CodonMPNN for Organism Specific and Codon Optimal Inverse Folding

Cameron Diao, Hannes Stark, Umesh Padia, Julia Balla, George Church

arXiv · PDF · Open in the Atlas

Code that ran

10 of 12 samples (functions and classes) lifted from this paper's linked repositories executed on a synthesized input in Syntology's jail (0 of them constructed an object rather than computing a result; 0 honoured the contract we drafted, 0 violated it and 0 have no contract verdict; 10 failures of our instrument; 2 no recorded run), measured 2026-11-22; 3 contract verdicts on these samples were measured on the host before the contract probe was jailed (2026-09-18T14:29:13Z) or are undated; repository links from the Papers with Code archive attributed under CC BY-SA, no affiliation; bodies inline only under a permissive licence as GitHub detects it. Not a reproduction of the paper's results.

"Ran" means the function executed on a synthesized input and returned a value. It is not a reproduction of the paper's results. This record, dated, as JSON: /api/ran/2409.17265.json · badge: /api/ran/2409.17265.svg

Each repository's role is a label: a classifier's output recorded on the link, or, where no role is recorded, what the Papers with Code archive marked. The line under it names its basis. Neither says who wrote the code.

RepositoryRoleRan
hannesstark/codonmpnn canonical (classifier label)
classifier basis: census
7 ran · 2 no recorded run
HannesStark/CodonMPNN canonical (classifier label)
classifier basis: classifier_2025
3 ran · 0 no recorded run
FunctionStatusWhere it lives
gather_edges Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
HannesStark/CodonMPNN/codon/utils/pmpnn.py
code served (permissive licence) · get_code("ee91093a4bcd2f9f")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1) and returned Tensor, ndim 4, dtype family float, shape (2, 3, 6, 4) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import numpy as np

torch.manual_seed(0)

# Analyze the function:
# gather_edges(edges, neighbor_idx)
# - neighbor_idx is unsqueezed to add a dimension, then expanded
# - edges is gathered along dimension 2 using neighbor_idx
# - edges.size(-1) is used to expand neighbor_idx
#
# So edges should be a 3D or 4D tensor where:
#   - dimension 2 is the dimension being gathered from
#   - dimension -1 (last) is the feature dimension
#
# neighbor_idx should be a 3D tensor that will be expanded to 4D

# Create edges tensor: shape (batch, nodes, num_edges, features)
# Using distinct sizes: batch=2, nodes=3, num_edges=5, features=4
edges = torch.randn(2, 3, 5, 4)

# Create neighbor_idx tensor: shape (batch, nodes, num_neighbors)
# This will be unsqueezed to (batch, nodes, num_neighbors, 1)
# then expanded to (batch, nodes, num_neighbors, features)
# Values should be valid indices for dimension 2 of edges (0-4)
neighbor_idx = torch.randint(0, 5, (2, 3, 6))

ARGS = {
    "edges": edges,
    "neighbor_idx": neighbor_idx,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
Imports
numpy, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Tensor, ndim 4, dtype family float, shape (2, 3, 6, 4) — shape (2, 3, 6, 4) as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
gather_log Ran
ran; our contract draft was wrong, not the code
hannesstark/codonmpnn/codon/wrapper.py
code served (permissive licence) · get_code("d1e5df8af4e52e48")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned dict. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import torch.nn as nn

torch.manual_seed(0)

# For gather_log, we need:
# 1. log: a dict where values are lists (since they get summed with [])
# 2. world_size: an integer

# Create a sample log dict with list values
# The function iterates over keys and sums lists, so values must be lists
log = {
    "loss": [0.5, 0.3],
    "accuracy": [0.9, 0.85],
    "steps": [100, 200]
}

world_size = 1

ARGS = {
    "log": log,
    "world_size": world_size
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type dict — a dict output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a dict output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
gather_nodes Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
HannesStark/CodonMPNN/codon/utils/pmpnn.py
code served (permissive licence) · get_code("2b89e352a944b025")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1) and returned Tensor, ndim 4, dtype family float, shape (2, 3, 2, 4) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import numpy as np

torch.manual_seed(0)

# Create nodes tensor: shape (batch_size, num_nodes, feature_dim)
# Using distinct dimensions: batch=2, num_nodes=5, feature_dim=4
nodes = torch.randn(2, 5, 4)

# Create neighbor_idx tensor: shape (batch_size, num_neighbors, k)
# Using distinct dimensions: batch=2, num_neighbors=3, k=2
neighbor_idx = torch.randint(0, 5, (2, 3, 2))

ARGS = {
    "nodes": nodes,
    "neighbor_idx": neighbor_idx,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
Imports
numpy, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Tensor, ndim 4, dtype family float, shape (2, 3, 2, 4) — shape (2, 3, 2, 4) as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
gather_nodes Ran
ran; our contract draft was wrong, not the code
hannesstark/codonmpnn/codon/utils/pmpnn.py
code served (permissive licence) · get_code("77cd7ee3a9fa2105")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned Tensor, ndim 4, dtype family float, shape (2, 3, 2, 4) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch

torch.manual_seed(0)

# nodes: shape (batch_size, num_nodes, feature_dim)
# neighbor_idx: shape (batch_size, num_neighbors, k) where k is some neighborhood size
nodes = torch.randn(2, 5, 4)  # batch=2, num_nodes=5, features=4

# neighbor_idx should contain indices into the nodes dimension (dim 1)
# shape: (batch_size, num_neighbors, k)
neighbor_idx = torch.randint(0, 5, (2, 3, 2))  # batch=2, num_neighbors=3, k=2

ARGS = {
    "nodes": nodes,
    "neighbor_idx": neighbor_idx,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Tensor, ndim 4, dtype family float, shape (2, 3, 2, 4) — shape (2, 3, 2, 4) as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
get_aligned_rmsd Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
hannesstark/codonmpnn/codon/utils/foldability_utils.py
code served (permissive licence) · get_code("80c136deefde6bce")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned float64, ndim 0, dtype family float; no output shape is stated. The image it ran in was not recorded. If your copy returns the same type and ndim on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import numpy as np

# The function get_aligned_rmsd expects two positional arguments: pos_1 and pos_2
# Both should be numpy arrays representing 3D positions
# Looking at the function body:
# - pos_1 and pos_2 are passed to rigid_transform_3D
# - rigid_transform_3D expects A and B with shape (N, 3) after transpose operations
# - So pos_1 and pos_2 should be (N, 3) shaped arrays of 3D coordinates

# Create small sample 3D point clouds
# Shape (5, 3) means 5 points in 3D space
pos_1 = np.array([
    [1.0, 2.0, 3.0],
    [4.0, 5.0, 6.0],
    [7.0, 8.0, 9.0],
    [10.0, 11.0, 12.0],
    [13.0, 14.0, 15.0]
], dtype=np.float64)

# Create a slightly transformed version of pos_1
# Apply a small rotation and translation to create pos_2
pos_2 = np.array([
    [1.1, 2.1, 3.1],
    [4.2, 5.2, 6.1],
    [7.1, 8.2, 9.0],
    [10.0, 11.1, 12.2],
    [13.1, 14.0, 15.1]
], dtype=np.float64)

ARGS = {
    'pos_1': pos_1,
    'pos_2': pos_2
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type float64, ndim 0, dtype family float — no output shape is stated: observe_signature.py recorded float ndim 0, which disagrees with the fixture run
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
get_log_mean Ran
ran; our contract draft was wrong, not the code
hannesstark/codonmpnn/codon/wrapper.py
code served (permissive licence) · get_code("6d7a23db92232fe7")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned dict. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import numpy as np

# Construct a log dictionary with various types of entries
# The function iterates through keys and tries to compute nanmean on values
log = {
    "loss": np.array([1.5, 2.3, np.nan, 3.1, 2.8]),
    "accuracy": np.array([0.85, 0.87, 0.86, np.nan, 0.88]),
    "val_loss": np.array([2.1, 2.4, 2.2]),
    "learning_rate": np.array([0.001, 0.001, 0.0009]),
    "non_numeric": ["a", "b", "c"],  # This will fail nanmean and be caught by except
    "empty_array": np.array([]),
}

ARGS = {
    "log": log
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type dict — a dict output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a dict output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
get_weird_pmpnn_stuff Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
HannesStark/CodonMPNN/codon/utils/pmpnn.py
code served (permissive licence) · get_code("e066a651617bbd2d")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1) and returned tuple, ndim 2, dtype family int; no output shape is stated. The image it ran in was not recorded. If your copy returns the same type and ndim on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import numpy as np

torch.manual_seed(0)

# Analyze the function: get_weird_pmpnn_stuff takes chain_idx as input
# chain_idx is used with:
# - len(chain_idx) to get L
# - torch.unique(chain_idx, return_counts=True) to get unique values and counts
# - indexing operations with count_cumsum

# Create a small chain_idx tensor
# It should be a 1D tensor of integers representing chain assignments
chain_idx = torch.tensor([0, 0, 1, 1, 2], dtype=torch.long)

ARGS = {
    "chain_idx": chain_idx
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
Imports
numpy, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple, ndim 2, dtype family int — a tuple output has no array shape of its own; none is stated
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
get_weird_pmpnn_stuff Ran
ran; our contract draft was wrong, not the code
hannesstark/codonmpnn/codon/utils/pmpnn.py
code served (permissive licence) · get_code("c7d3a02c3dc97e6a")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned tuple, ndim 2, dtype family int; no output shape is stated. The image it ran in was not recorded. If your copy returns the same type and ndim on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch

torch.manual_seed(0)

# chain_idx is a tensor of chain identifiers
# The function expects it to be a 1D tensor of integers
# Let's create a small chain_idx with multiple chains
chain_idx = torch.tensor([0, 0, 0, 1, 1, 2], dtype=torch.long)

ARGS = {
    "chain_idx": chain_idx,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple, ndim 2, dtype family int — a tuple output has no array shape of its own; none is stated
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
multi_seq_collate Ran
ran; our contract draft was wrong, not the code
hannesstark/codonmpnn/codon/datasets.py
code served (permissive licence) · get_code("5693967e08c16256")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned dict. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
from torch.utils.data import default_collate

torch.manual_seed(0)

# Create a batch of items where each item is a dict with the required keys
batch = []

# Item 1
item1 = {
    'atom37': torch.randn(5, 4),  # sequence length 5, feature dim 4
    'seq': torch.randn(5, 3),     # sequence length 5, feature dim 3
    'atom_mask': torch.ones(5, 2), # sequence length 5, feature dim 2
    'wildtype_codons': torch.randint(0, 10, (5, 4)), # sequence length 5
    'mut_codons': torch.randint(0, 10, (5, 4)),  # sequence length 5
    'pmpnn_res_idx': torch.arange(5).unsqueeze(1),  # sequence length 5
    'pmpnn_chain_encoding': torch.ones(5, 1),  # sequence length 5
    'other_data': torch.randn(2, 3),  # additional data not in seq_len_keys
    'scalar_value': 42,  # scalar data for default_collate
}

# Item 2 (shorter sequence)
item2 = {
    'atom37': torch.randn(3, 4),   # sequence length 3, feature dim 4
    'seq': torch.randn(3, 3),      # sequence length 3, feature dim 3
    'atom_mask': torch.ones(3, 2), # sequence length 3, feature dim 2
    'wildtype_codons': torch.randint(0, 10, (3, 4)), # sequence length 3
    'mut_codons': torch.randint(0, 10, (3, 4)),  # sequence length 3
    'pmpnn_res_idx': torch.arange(3).unsqueeze(1),  # sequence length 3
    'pmpnn_chain_encoding': torch.ones(3, 1),  # sequence length 3
    'other_data': torch.randn(2, 3),  # additional data not in seq_len_keys
    'scalar_value': 99,  # scalar data for default_collate
}

batch = [item1, item2]

ARGS = {
    'batch': batch,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type dict — a dict output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a dict output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
seq_collate Ran
ran; our contract draft was wrong, not the code
hannesstark/codonmpnn/codon/datasets.py
code served (permissive licence) · get_code("76963c496a2e0091")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned dict. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
from torch.utils.data import default_collate

torch.manual_seed(0)

# Create a batch of items following the expected structure
# Each item is a dict with seq_len_keys and potentially other keys

batch = []

# Item 1
item1 = {
    'atom37': torch.randn(5, 4),  # L=5, feature_dim=4
    'seq': torch.randint(0, 20, (5, 4)),  # L=5, feature_dim=4
    'atom_mask': torch.ones(5, 4),  # L=5, feature_dim=4
    'codons': torch.randint(0, 4, (5, 4)),  # L=5, feature_dim=4
    'pmpnn_res_idx': torch.arange(5).unsqueeze(1).expand(5, 4),  # L=5, feature_dim=4
    'pmpnn_chain_encoding': torch.zeros(5, 4),  # L=5, feature_dim=4
    'extra_field': torch.randn(2, 3),  # Extra field to test default_collate
}
batch.append(item1)

# Item 2 (shorter sequence)
item2 = {
    'atom37': torch.randn(3, 4),  # L=3, feature_dim=4
    'seq': torch.randint(0, 20, (3, 4)),  # L=3, feature_dim=4
    'atom_mask': torch.ones(3, 4),  # L=3, feature_dim=4
    'codons': torch.randint(0, 4, (3, 4)),  # L=3, feature_dim=4
    'pmpnn_res_idx': torch.arange(3).unsqueeze(1).expand(3, 4),  # L=3, feature_dim=4
    'pmpnn_chain_encoding': torch.zeros(3, 4),  # L=3, feature_dim=4
    'extra_field': torch.randn(2, 3),  # Extra field to test default_collate
}
batch.append(item2)

ARGS = {
    'batch': batch,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type dict — a dict output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a dict output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
gather_edges Not yet run
unverified (harvested, no recorded run)
hannesstark/codonmpnn/codon/utils/pmpnn.py
code served (permissive licence) · get_code("ae83abf1a9db1e77")
How it ran

no invocation recorded.

get_logger Not yet run
unverified (harvested, no recorded run)
hannesstark/codonmpnn/codon/utils/logging.py
code served (permissive licence) · get_code("ec9a324c17c1d5ce")
How it ran

no invocation recorded.

Also present in other papers' code

Of 12 samples lifted from this paper's linked repositories, 12 carry a function identity and 0 do not. Of the 9 distinct functions with an identity, 4 are also present in other papers whose code we hold (4 with at least one paper dated earlier, 1 the same month, 1 later) and 5 are only here among the 52,977 papers whose code Syntology holds (as of 2026-09-26).

Only here means: among the 52,977 papers whose code Syntology holds (51,887 with lifted samples, 1,090 through their assemblies only), as of 2026-09-26, no other paper's lifted samples or current assembly bricks contain an identical function body after normalisation.

FunctionAlso present inEarliest-datedOthers named
gather_log 2 other papers
2 dated earlier
2310.05764 (2023-10)
Harmonic Self-Conditioned Flow Matching for joint Multi-Ligand Docking and Binding Site Design
2402.04845 (2024-02)
AlphaFold Meets Flow Matching for Generating Protein Ensembles
gather_nodes 2 other papers
1 dated earlier, 1 later
2110.04624 (2021-10)
Iterative Refinement Graph Neural Network for Antibody Sequence-Structure Co-design
2601.19232 (2026-01)
Structure-based RNA Design by Step-wise Optimization of Latent Diffusion Model
get_logger 2 other papers
1 dated earlier, 1 the same month
2402.04845 (2024-02)
AlphaFold Meets Flow Matching for Generating Protein Ensembles
2409.17808 (2024-09)
Generative Modeling of Molecular Dynamics Trajectories
gather_edges 1 other paper
1 dated earlier
2110.04624 (2021-10)
Iterative Refinement Graph Neural Network for Antibody Sequence-Structure Co-design

Presence, not direction: an identical function body in two papers' repositories does not say who wrote it, which came from which, or that either paper's method rests on it. Matching is exact after normalisation, so a renamed or edited function is not matched. Dates are when papers appeared, not when code was written. Only .py files the harvester lifted are covered; notebooks are not read. A new-style arXiv id YYMM.NNNN dates a paper to the month 20YY-MM; otherwise Paper.year dates it to the year; otherwise it is undated. Dates are when the paper appeared, not when the code was written. Read 2026-11-22.

Repositories linked to this paper

Some links come from the archived Papers with Code dataset (CC BY-SA 4.0): attribution and licence.

Abstract

Generating protein sequences conditioned on protein structures is an impactful technique for protein engineering. When synthesizing engineered proteins, they are commonly translated into DNA and expressed in an organism such as yeast. One difficulty in this process is that the expression rates can be low due to suboptimal codon sequences for expressing a protein in a host organism. We propose CodonMPNN, which generates a codon sequence conditioned on a protein backbone structure and an organism label. If naturally occurring DNA sequences are close to codon optimality, CodonMPNN could learn to generate codon sequences with higher expression yields than heuristic codon choices for generated amino acid sequences. Experiments show that CodonMPNN retains the performance of previous inverse folding approaches and recovers wildtype codons more frequently than baselines. Furthermore, CodonMPNN has a higher likelihood of generating high-fitness codon sequences than low-fitness codon sequences for the same protein sequence. Code is available at https:// github.com/HannesStark/CodonMPNN.

For agents

The same record, over MCP at https://syntology.ai/mcp:

get_harvested_code_for_paper("2409.17265")
get_code_for_paper("2409.17265")
have("2409.17265")

The run record, dated, one paper per request, free:

curl https://syntology.ai/api/ran/2409.17265.json

A badge for a README (the split and the date, never a ratio):

[![Syntology run record](https://syntology.ai/api/ran/2409.17265.svg)](https://syntology.ai/paper/2409.17265)

Connect an agent — have() is free.