Maps likely protein-protein interactions from amino acid sequence alone.

Atlas PPI

Atlas PPI maps likely protein-protein interactions from amino acid sequence alone. It is the production-facing interaction model behind Synthyra's Atlas platform and builds on the Synteract research line.

Atlas PPI is not a simple port of an open-source model. Synthyra products extend open research with additional model development, calibration, serving infrastructure, validation workflows, and analysis layers so researchers can move from a sequence to an actionable interaction map.

What It Does

Atlas PPI helps researchers:

• Find likely interaction partners for a protein of interest.
• Expand an interaction neighborhood into a network.
• Screen a query protein against a reference proteome.
• Compare designed proteins against natural proteins before synthesis.
• Identify possible off-target interaction risks.
• Prioritize experiments by ranking plausible interaction partners.

The output is a prioritization signal. It helps decide which biological hypotheses deserve a closer look.

Why It Matters

Protein-protein interactions are the wiring of the cell, but measuring them directly is expensive and incomplete. Atlas PPI gives researchers a fast first-pass map of that wiring.

The model can support drug discovery, target biology, protein design, synthetic biology, and safety screening by helping teams narrow millions of possible protein pairs into a smaller set of high-value experiments.

The Synteract-4 research behind Atlas PPI beats the best published comparator by a 13% margin on a gold-standard PPI benchmark and recovers many newly measured HSP90 interaction partners that were missing from prior databases. Those results are important because they show that sequence-only interaction mapping can be useful beyond simply repeating known biology.

Open Research Foundation

Atlas PPI is based on the Synteract-4 research direction: using protein sequences to build interaction maps at proteome scale. Synthyra's product version extends that foundation into a practical system for real workflows, including API access, proteome-scale analysis, network generation, cross-reference views, and integration with the broader Atlas platform.

This distinction matters. The open research establishes the scientific foundation. The Synthyra product turns that foundation into a continuously improved, usable platform capability.

Using Atlas PPI

Score matched protein pairs over HTTPS with your Synthyra API key:

import requests

api_key = "..."  # synthyra.com/settings?section=api-keys

resp = requests.post(
    "https://api.synthyra.com/v1/score/pairs",
    headers={"Authorization": f"Bearer {api_key}"},
    json={
        "inputs_a": ["MEV...", "MAK..."],
        "inputs_b": ["MQT...", "MLG..."],
        "ids_a": ["P04637", "P12345"],
        "ids_b": ["P38398", "Q9Y6K9"],
        "both_directions": True,
    },
    timeout=180,
)
result = resp.json()  # {"scores": [...], "ids_a": [...], "ids_b": [...]}

/v1/score/pairs returns quantized confidence scores in [0, 100]. For all-vs-all matrices use /v1/score/matrix; for network expansion around query proteins use /v1/generate/network. Pricing: $0.0001 per pair.

Limitations

Atlas PPI should be interpreted as a predictive research tool, not a source of ground truth.

• Scores do not measure binding affinity.
• Scores do not prove direct physical contact.
• Cellular context, post-translational modification, expression level, localization, and cofactors can change whether an interaction occurs in vivo.
• Host-pathogen and cross-species interactions remain harder than same-species interaction maps.
• Novel designed proteins can fall outside the natural training distribution and should be validated experimentally.

For best results, combine Atlas PPI scores with orthogonal evidence such as STRING, BioGRID, structural modeling, localization, expression context, and wet-lab validation.