Search for similar proteins

Use sequence similarity search when you have an amino-acid sequence and want Atlas proteins with related SAE feature profiles. This is feature similarity, not a sequence-alignment or homology search.

Before you begin

Prepare a valid protein sequence of at most 2,048 residues. Review the similarity-search endpoint for all filters, limits, and response fields.

Submit the sequence

This example requests five results and includes cluster summaries.

import httpx

sequence = "FVNQHLCGSHLVEALYLVCGERGFFYTPKT"
response = httpx.get(
    "https://biohub.ai/esm/protein/api/v1alpha1/similarity-search",
    params={
        "sequence": sequence,
        "topk_results": 5,
        "topk_features": 20,
        "include_cluster_info": True,
    },
    timeout=60,
)
response.raise_for_status()
result = response.json()

Inspect every returned protein

for protein in result["similar_proteins"]:
    print(
        protein["protein_accession"],
        protein["similarity_score"],
        protein.get("cluster_size"),
    )

Results are ordered by descending similarity score. Preserve each returned protein_hash when you need to retrieve the complete protein record.

Check withheld-result metadata

print("Restricted results:", result["restricted_count"])
print("Shared top features:", result["top_features_across_results"])

restricted_count reports results withheld by the Atlas biosecurity filter. Do not infer the identity or content of withheld proteins.

A successful request returns zero or more feature-similar proteins plus aggregate feature evidence. An empty result is a valid outcome and should not be broadened automatically without reviewing the query and filters.

Next step

Pass a returned protein_hash to Retrieve protein and cluster context, or inspect the exact protein endpoint contract.

​Before you begin

​Next step

Before you begin

Next step