SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2603.20955 · 2026

Beyond Expression Similarity: Contrastive Learning Recovers Functional Gene Associations from Protein Interaction Structure

Jason Dury

arXiv · PDF · Open in the Atlas

Code that ran

7 of 7 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; 7 failures of our instrument; 0 no recorded run), measured 2026-11-22; 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/2603.20955.json · badge: /api/ran/2603.20955.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
EridosAI/GeneticCAL canonical (classifier label)
classifier basis: census
7 ran · 0 no recorded run
FunctionStatusWhere it lives
classify_cluster Ran
ran; our contract draft was wrong, not the code
EridosAI/GeneticCAL/paper/strengthening_results/run_05_expression_enrichment.py
code served (permissive licence) · get_code("529b8fd5dae0a74c")
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 str. 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

ARGS = {
    'enrichments': [
        {'p_value': 0.02, 'coverage': 0.6},
        {'p_value': 0.10, 'coverage': 0.3}
    ]
}
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 str — a str output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a str 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
compute_scores Ran
ran; our contract draft was wrong, not the code
EridosAI/GeneticCAL/paper/strengthening_results/run_01_node_split.py
code served (permissive licence) · get_code("21db26aff31a21a8")
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 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 torch
import torch.nn as nn

torch.manual_seed(0)

# Define a minimal model that satisfies the function's usage:
# - model.eval() is called
# - model(emb_a) and model(emb_b) are called with tensors of shape (batch, dim)
# - output should be a tensor of shape (batch, dim) for dot products with embeddings
class MinimalModel(nn.Module):
    def __init__(self, input_dim, output_dim):
        super().__init__()
        self.linear = nn.Linear(input_dim, output_dim)
    
    def forward(self, x):
        return self.linear(x)

# Create model
model = MinimalModel(input_dim=4, output_dim=4)

# Create normalized embedding tensor
# Shape: (num_embeddings, embedding_dim) = (6, 4)
X_norm_tensor = torch.randn(6, 4)
X_norm_tensor = X_norm_tensor / (X_norm_tensor.norm(dim=-1, keepdim=True) + 1e-8)

# Create index tensors for selecting embeddings
# idx_a and idx_b should be 1D tensors with indices into X_norm_tensor
idx_a = torch.tensor([0, 1])
idx_b = torch.tensor([2, 3])

ARGS = {
    "model": model,
    "X_norm_tensor": X_norm_tensor,
    "idx_a": idx_a,
    "idx_b": idx_b,
}
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 float — 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
compute_scores Ran
ran; our contract draft was wrong, not the code
EridosAI/GeneticCAL/paper/strengthening_results/run_02_seed_variance.py
code served (permissive licence) · get_code("47d2aa08855bb1e7")
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 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 torch
import torch.nn as nn

torch.manual_seed(0)

# Define a simple model that takes embeddings and produces predictions
class SimpleModel(nn.Module):
    def __init__(self, input_dim, output_dim):
        super().__init__()
        self.linear = nn.Linear(input_dim, output_dim)
    
    def forward(self, x):
        return self.linear(x)

# Create model
model = SimpleModel(input_dim=4, output_dim=4)

# Create normalized embedding tensor: batch_size=2, embedding_dim=4
X_norm_tensor = torch.randn(2, 4)
X_norm_tensor = X_norm_tensor / (X_norm_tensor.norm(dim=-1, keepdim=True) + 1e-8)

# Create indices for selecting embeddings
idx_a = torch.tensor([0])
idx_b = torch.tensor([1])

ARGS = {
    "model": model,
    "X_norm_tensor": X_norm_tensor,
    "idx_a": idx_a,
    "idx_b": idx_b,
}
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 float — 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
contrastive_loss Ran
ran; our contract draft was wrong, not the code
EridosAI/GeneticCAL/paper/strengthening_results/run_01_node_split.py
code served (permissive licence) · get_code("cffe9ba22ca505ad")
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 0, dtype family float, shape () 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 torch.nn.functional as F

torch.manual_seed(0)

# Create small embedding tensors
# emb_a and emb_b should have the same batch size and embedding dimension
batch_size = 2
embedding_dim = 4

emb_a = torch.randn(batch_size, embedding_dim)
emb_b = torch.randn(batch_size, embedding_dim)

temperature = 0.05

ARGS = {
    "emb_a": emb_a,
    "emb_b": emb_b,
    "temperature": temperature,
}
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 0, dtype family float, shape () — shape () 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
train_model Ran
ran; our contract draft was wrong, not the code
EridosAI/GeneticCAL/paper/strengthening_results/run_01_node_split.py
code served (permissive licence) · get_code("0082b3491c09c97e")
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 1, dtype family float; no output shape is stated. The image it ran in was not recorded. The code mentions cuda, and the harness aliases cuda to the CPU for every sample; whether this run needed that is 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 torch.nn as nn
import numpy as np

torch.manual_seed(0)
np.random.seed(0)

# Create minimal X_norm_tensor: shape (10, 50) - 10 samples, 50 PCA dimensions
X_norm_tensor = torch.randn(10, 50)

# Create pairs: shape (8, 2) - 8 pairs of indices into X_norm_tensor
pairs = np.array([
    [0, 1],
    [1, 2],
    [2, 3],
    [3, 4],
    [4, 5],
    [5, 6],
    [6, 7],
    [7, 8]
], dtype=np.int64)

ARGS = {
    'X_norm_tensor': X_norm_tensor,
    'pairs': pairs,
    'epochs': 2
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
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 1, dtype family float — 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
train_model Ran
ran; our contract draft was wrong, not the code
EridosAI/GeneticCAL/paper/strengthening_results/run_02_seed_variance.py
code served (permissive licence) · get_code("eb0e70cfaaad0fa1")
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. The image it ran in was not recorded. The code mentions cuda, and the harness aliases cuda to the CPU for every sample; whether this run needed that is 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
import torch
import torch.nn as nn

torch.manual_seed(0)
np.random.seed(0)

# Create small dummy tensors
# X_norm_tensor: shape (n_samples, PCA_DIM=50) - normalized embeddings
n_samples = 16
X_norm_tensor = torch.randn(n_samples, 50)
X_norm_tensor = X_norm_tensor / (torch.norm(X_norm_tensor, dim=1, keepdim=True) + 1e-8)
X_norm_tensor = X_norm_tensor.to('cpu')

# pairs: shape (n_pairs, 2) - indices into X_norm_tensor
# Each row is [idx_a, idx_b] for a positive pair
pairs = np.array([
    [0, 1],
    [1, 2],
    [2, 3],
    [3, 4],
    [4, 5],
    [5, 6],
    [6, 7],
    [7, 8],
    [8, 9],
    [9, 10],
    [10, 11],
    [11, 12],
    [12, 13],
    [13, 14],
    [14, 15],
    [15, 0],
], dtype=np.int64)

# seed parameter
seed = 42

# epochs parameter
epochs = 2

ARGS = {
    "X_norm_tensor": X_norm_tensor,
    "pairs": pairs,
    "seed": seed,
    "epochs": epochs,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
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 — a tuple output has no array shape of its own; none is 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
train_model Ran
ran; our contract draft was wrong, not the code
EridosAI/GeneticCAL/paper/strengthening_results/run_03_experimental_string.py
code served (permissive licence) · get_code("64674e3a42b3a7d1")
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 1, dtype family float; no output shape is stated. The image it ran in was not recorded. The code mentions cuda, and the harness aliases cuda to the CPU for every sample; whether this run needed that is 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
import torch
import torch.nn as nn

torch.manual_seed(0)
np.random.seed(0)

# Create tiny normalized embeddings tensor
# Shape: (10, 50) - 10 samples, 50 PCA dimensions
X_norm_tensor = torch.randn(10, 50)
X_norm_tensor = X_norm_tensor / (torch.norm(X_norm_tensor, dim=1, keepdim=True) + 1e-8)
X_norm_tensor = X_norm_tensor.to(torch.float32)

# Create pairs as indices into X_norm_tensor
# Shape: (16, 2) - 16 pairs, each pair has 2 indices
pairs = np.array([
    [0, 1], [1, 2], [2, 3], [3, 4],
    [4, 5], [5, 6], [6, 7], [7, 8],
    [8, 9], [9, 0], [1, 3], [2, 5],
    [3, 7], [4, 8], [5, 9], [6, 0]
], dtype=np.int64)

ARGS = {
    "X_norm_tensor": X_norm_tensor,
    "pairs": pairs,
    "epochs": 2
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
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 1, dtype family float — 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

Also present in other papers' code

Of 7 samples lifted from this paper's linked repositories, 7 carry a function identity and 0 do not. Of the 6 distinct functions with an identity, 0 are also present in other papers whose code we hold (0 with at least one paper dated earlier, 0 the same month, 0 later) and 6 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.

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

The Predictive Associative Memory (PAM) framework posits that useful relationships often connect items that co-occur in shared contexts rather than items that appear similar in embedding space. 1 A contrastive MLP trained on co-occurrence annotations-the procedure we term Contrastive Association Learning (CAL)-has been shown to improve multi-hop passage retrieval and discover narrative function at corpus scale in text. We test whether this principle transfers to molecular biology, where protein-protein interactions provide independently derived functional associations distinct from gene expression similarity. Four experiments across two biological domains map the operating envelope. On gene perturbation data (Replogle K562 CRISPRi, 2,285 genes), CAL trained on STRING protein interactions achieves cross-boundary AUC of 0.908 where expression similarity scores 0.518. A second gene dataset (DepMap, 17,725 genes) confirms the result after a negative sampling correction, reaching cross-boundary AUC of 0.947. Two drug sensitivity experiments produce informative negatives that sharpen the boundary conditions. Three cross-domain findings emerge: (1) inductive transfer succeeds in biology-a node-disjoint split with completely unseen genes yields AUC 0.826 (∆ +0.127)-where it fails in text (±0.10), suggesting physically grounded associations are more transferable than contingent co-occurrences; (2) CAL association scores anti-correlate with interaction degree (Spearman r = -0.590), with cross-boundary gains concentrating on understudied genes with focused interaction profiles; (3) tighter association quality outperforms larger but noisier training sets, reversing the pattern observed in text. Results on the primary dataset are stable across training seeds (SD < 0.001) and cross-boundary threshold choices.

For agents

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

get_harvested_code_for_paper("2603.20955")
get_code_for_paper("2603.20955")
have("2603.20955")

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

curl https://syntology.ai/api/ran/2603.20955.json

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

[![Syntology run record](https://syntology.ai/api/ran/2603.20955.svg)](https://syntology.ai/paper/2603.20955)

Connect an agent — have() is free.