Receptor.AI Opens Access to Its End-to-End Peptide Discovery Platform

Cambridge, MA — 14 October, 2026 — Receptor.AI, a drug discovery company developing structure-guided design platforms, today opened access to its End-to-End Peptide Discovery Platform. External research teams can now use the technology developed through the company's own peptide discovery programs to design and optimize therapeutic peptides.

The platform supports hit identification, peptide–target complex modeling and affinity optimization, as well as passive membrane permeability prediction and multiparameter design. The first SaaS release, available from October 14, provides structure-guided affinity optimization. Other capabilities are currently available through collaborative discovery programs and are planned for future SaaS releases. An accompanying white paper presents the platform's methodology and validation results, with four supporting benchmark datasets released alongside it.

Therapeutic peptide design presents several interconnected challenges. Optimization often starts with evidence that a peptide binds to a target, but limited information about how the two interact. Even a single amino-acid substitution can change the peptide's shape and its interactions with the target, strengthening or weakening binding. Understanding these structural effects helps researchers decide which modifications to investigate before committing to synthesis.

The platform uses three-dimensional representations of peptide–target complexes to guide optimization, with capabilities that address these challenges in peptide design:

ForceFold models how a peptide binds by combining a physics-based search across receptor states, peptide conformations, and docking poses with an all-atom co-folding model guided by proprietary conditioning adaptors. When several binding modes remain plausible, ForceFold retains those alternatives for further evaluation.

The platform also supports a wider range of peptide chemistry. Non-canonical residues, including D-isomers, N-methylated residues and beta- and gamma-amino acids, are parameterized in advance together with cross-linker attachment groups and backbone modifications. Researchers can therefore use them in conformational sampling, docking, scoring and molecular dynamics without configuring them separately for each project.

Affinity and permeability can be evaluated within the same workflow. The platform ranks candidates by relative affinity in the modeled complex and assesses passive membrane permeability for the same candidates. Permeability prediction combines a conformation-sensitive AI model with two physics-based tiers for candidates that the AI model cannot assess confidently.

Starting from an existing peptide and a defined set of permitted substitutions, QuorumMap searches the resulting chemical space and prioritizes variants for further computation and synthesis. StratAI coordinates the broader design–make–test–analyze cycle, using laboratory results to guide subsequent optimization rounds.

“Research teams need a practical way to decide which peptide designs are worth taking into the lab,” said Alan Nafiev, Ph.D., Chief Executive Officer and Founder of Receptor.AI. “We have spent several years developing and applying this technology in our own discovery programs. Opening access gives other teams the same tools to guide their optimization work and explore a much broader range of peptide chemistry.”

The white paper reports benchmarks for individual platform components. ForceFold achieved CAPRI Acceptable-or-better accuracy for 74.4% of top-ranked predictions across 172 protein–peptide complexes and outperformed the same co-folding model supplied with physical contacts through its native restraint input.

The permeability AI model achieved a Spearman correlation of 0.71 on held-out scaffolds, compared with 0.68 for a published reference model. For peptide lengths withheld from training and lengths beyond the training range, correlations were 0.64 and 0.59, respectively, compared with 0.46 and 0.33 for the reference model.

In an experimental affinity optimization program, a single computational pass produced 24 designs for synthesis, nine of which bound with affinity below 100 nM.

The experimental programs did not include matched alternative workflows, so these results do not establish reductions in synthesis requirements, design cycles, or development time.

“Experimental screening gives us peptide hits. Identifying how those hits bind to their target gives us a starting point for structure-guided optimization,” said Semen Yesylevskyy, Ph.D., Chief Scientific Officer and Co-founder of Receptor.AI. “We connect binding mode identification directly to the design of new variants, using the peptide–target interactions to guide which positions to modify and which substitutions to explore. This makes an experimental hit the starting point for a systematic search through a much broader chemical space.”

Access is available through SaaS subscriptions, commercial licenses, and collaborative discovery agreements.

‍

About Receptor.AI

Receptor.AI develops therapeutic peptides and an end-to-end platform for their design and optimization. The company combines expertise in peptide chemistry with AI, physics-based modeling, and experimental validation to advance its own discovery programs and those of its partners.