Biotech · 10 min read

Design-Make-Test-Learn: Receptor.AI and Sethera Close the Loop on Polymacrocyclic Peptides

Receptor.AI and Sethera Therapeutics announced an August 3 alliance pairing encoded polymacrocyclic peptide libraries with physics-based AI for a design-make-test-learn workflow on hard-to-drug targets.

By Classy AI News Staff — Research Desk · August 4, 2026

AI drug discovery pitches often stop at prediction. The harder problem is closing the loop: generate a candidate, test it, learn from the result, and iterate — fast enough that the model improves with every cycle rather than every funding round.

On August 3, 2026, Receptor.AI and Sethera Therapeutics announced a strategic collaboration to build exactly that workflow for polymacrocyclic peptide medicines targeting difficult therapeutic targets.

What each side brings

Sethera Therapeutics develops an enzymatic platform that installs one to six stable thioether cross-links to create polymacrocyclic, nested, in-line, and interpeptide structures across large encoded libraries. Unlike constrained-peptide approaches centered on a predetermined structural motif, Sethera's platform explores multiple experimentally accessible topologies.

Receptor.AI contributes physics-based modeling, generative AI, and multiparameter optimization to interpret sequence, architecture, enrichment, and activity data — developing binding hypotheses, prioritizing candidate series, and guiding focused optimization cycles.

Under the collaboration, Sethera generates and experimentally screens architecture-diverse libraries. Receptor.AI interprets the results. The companies then design, synthesize, and test new candidates, feeding data back into the next cycle.

Microscope and lab samples in a biotechnology research setting

Public statements

Dr. Alan Nafiiev, founder and CEO of Receptor.AI, said in the Newswise release:

"Sethera's platform creates experimentally accessible peptide architectures that conventional design approaches do not readily reach. Our objective is to use physics and AI not as a substitute for experimentation, but to learn from each experimental cycle and direct the next one. That closed feedback loop is where we believe the collaboration can create distinctive value."

Dr. Karsten Eastman, CEO and co-founder of Sethera Therapeutics, added:

"Receptor.AI adds a powerful layer for understanding why those hits work and how they can be improved. Together, we intend to create a coordinated design-make-test-learn process that moves more efficiently from experimental discovery to validated lead series."

The initial program

The first project focuses on a mutually selected hard-to-drug target. The companies will prospectively assess whether the integrated workflow improves hit confirmation, target selectivity, and lead optimization compared with conventional enrichment- and assay-led prioritization.

If validated, Receptor.AI and Sethera intend to pursue additional internal programs and jointly structured discovery collaborations with pharmaceutical and biotechnology partners. Additional terms were not disclosed.

GEN and FirstWord Pharma covered the announcement the same day.

Research scientist examining samples in a clinical laboratory

Why peptides, why now

Macrocyclic and polymacrocyclic peptides occupy a middle ground between small molecules and biologics — large enough for selective protein interfaces, sometimes small enough for oral delivery challenges to be engineering problems rather than dead ends.

The collaboration's stated goal is a repeatable discovery system that continuously learns from sequence, architecture, counterselection, binding, functional, and developability data across experimental campaigns — reducing design cycles from screening hits to differentiated lead series.

That is the same architectural bet GSK made in its expanded $110 million collaboration with Relation Therapeutics — pair data generation with model development — applied here to encoded peptide libraries rather than cellular perturbation screens.

What would count as proof

Vendor partnerships are easy to announce and hard to validate. Meaningful evidence would include:

  • Published lead series with disclosed potency, selectivity, and developability metrics versus conventional workflows on the same target.
  • Cycle-time reduction data: number of design-make-test iterations to reach a validated lead.
  • Partner expansion beyond the initial hard-to-drug program.

Until then, the August 3 announcement is a credible alignment of complementary platforms — not yet a demonstrated acceleration.

Glass vials and research tools arranged in a laboratory

### Sources