MoleculeMind MMDesign Technical Report: 18 of 19 Projects Successful, All 6 GPCR Targets Hit

18 of 19 projects succeeded, and all 6 GPCR targets were hit! Domestic AI drug discovery breaks the technical ceiling again. QbitAI Download the client for exclusive early access. October 28, 2026, 13:24:28 From Beijing Updated October 28, 2026, 16:41:20

QbitAI | WeChat public account QbitAI

A technical report recently released by MoleculeMind, however, gives a somewhat different answer.

This latest technical report on its AI de novo protein design platform MMDesign covers 19 design projects across 18 different targets and two molecular modalities, nanobodies (VHH) and macrocyclic peptides:

Each project tested only 9-50 candidate molecules; 18 projects yielded experimentally confirmed binding molecules, for a project success rate of 94.7%;

All six GPCR targets tested were successful, with best affinities all in the nanomolar range;

For six targets that had already been hit previously, redesign was performed, and the design results for five projects were further improved;

Some candidate molecules also showed promising preliminary developability and sequence novelty.

Together, these numbers point to MMDesign having moved from "technical breakthrough" into the "systematic validation" stage: the same "generate-screen-validate" workflow can repeatedly pass wet-lab validation with a small number of candidate molecules across targets and modalities.

MMDesign, which has undergone three iterations within six months, is trying to turn "efficient hits" into a repeatable capability.

After generation, can fewer candidate molecules be sent to the lab for testing, and can the hit rate be higher? This is efficiency.

Can the affinity, developability metrics, and novelty of candidate molecules be better? This is quality.

Another point is equally critical: the generation process does not rely on any known binding molecule as a reference for modification, but instead designs molecules directly from functional requirements, and can provide an answer even for targets with no precedent to follow. This is precisely the essential value of de novo design as distinct from "modifying known molecules."

First, move the "mass screening" into the computer: a single project yields hits by testing only a few dozen molecules

But generating sequences is only the starting point. Being able to precisely pick, from tens of thousands of candidates, the few dozen worth taking into the lab is the real dividing line.

MMDesign fixes this process into a unified "generate-screen-validate" workflow:

First, generate tens of thousands of candidates in computational space, then, through structural reliability, sequence naturalness, and multi-layer evaluation based on statistical physics, compress the candidate set for a single project to only 9-50, with a median of only 21. Of 19 projects, 18 obtained binding molecules, for a project success rate of 94.7%; among all 453 tested candidates, 139 were confirmed to bind, for an overall candidate hit rate of 30.7%.

There are two yardsticks here, different in definition but both important: 94.7% measures "whether a project can find at least one binding molecule," while 30.7% measures "how many of the tested candidates actually hit." For de novo design, both numbers are already quite impressive globally.

15 projects obtained at least two binding molecules, accounting for nearly 80% of all projects, meaning R&D teams not only "have something to use" but can also compare and optimize among multiple candidates.

For pharmaceutical companies, excluding low-value candidates as early as possible on the computational side allows experimental resources to be concentrated on high-potential molecules. This is the real value of moving from "mass screening" to "precision design."

Six GPCRs, not a single one failed

If 94.7% demonstrates efficiency, the GPCR series directly tests the technical caliber.

GPCRs are one of the most important drug target families in the pharmaceutical industry, but designing antibodies against them has long been difficult.

On the one hand, as membrane proteins, GPCRs have complex structures and variable conformations; on the other hand, the regions exposed outside the cell that are available for antibody binding are very limited. These characteristics make GPCRs a recognized "hard nut" in the field of de novo antibody design.

The report shows that MMDesign carried out de novo design against six GPCR targets—CCR7, CCR8, EP2, GPR68, NK3R, and A2AR—and all successfully obtained experimentally confirmed binding molecules.

Under the BLI assay conditions used in the report, candidate hit rates for each target ranged from 12.1% to 100%, and the best binding molecules for all six targets reached nanomolar affinity, with GPR68's best KD as low as 12.1 nanomolar.

Two sets of results are especially striking: all 30 tested candidates for NK3R were confirmed to bind, a hit rate of 100%; and 10 of 11 candidates for GPR68 hit.

Further evidence comes from a "no precedent" scenario:

On five GPCR targets for which a search of the Thera-SAbDab database (PL132) found no cataloged precedent therapeutic antibodies, 53 of 89 tested designs obtained binding, for a candidate hit rate of 59.6%.

The 59.6% candidate hit rate shows that even without an "answer key," the platform can still open new discovery paths on complex membrane protein targets, offering the possibility of breaking through traditional R&D bottlenecks.

Optimizable across rounds, extensible across modalities

What pharmaceutical companies and platforms truly value in long-term collaboration is "whether it can be used repeatedly": after an initial design, can it still be optimized? If the task changes, can it still succeed?

This report gives an affirmative answer: MMDesign performed another round of design on six targets that already had binding molecules, and five of them achieved further improvements in hit rate, best affinity, or both.

Taking PDGFR as an example, the candidate hit rate rose from 4.0% to 21.1%, and the best KD was optimized from 150 nanomolar to 37 nanomolar, corresponding to a 4-fold affinity improvement. For the same R&D problem, another round can still yield better molecules.

The platform's scope of application also extends to different molecular modalities: macrocyclic peptide design results show that in the MDM2 project, 16 of 21 candidates hit, a hit rate of 76.2%, with a best KD of 13.7 micromolar; in the PD-L1 project, 11 of 28 candidates hit, with a best KD of 40.2 micromolar.

Compared with nanobodies, the conformations of macrocyclic peptides are constrained by the cyclic backbone, so the design challenges differ, but it works just as well.

This "generation-screening-validation" capability is being reused for different design tasks.

Cross-round design optimization and cross-modality workflow reuse together constitute two sides of the same systematic validation:

The platform's capability is moving from "single-point success" toward "systematic output."

△Summary of validation results

"Being able to bind" is the entry ticket, but developability and novelty also need to pass muster

For drug discovery, "being able to bind" counts as getting the entry ticket, but a truly valuable early hit molecule still has to pass several subsequent "checkpoints":

whether it can be stably expressed, whether it will aggregate, and whether it binds unrelated proteins nonspecifically.

The report shows that MMDesign has already moved such issues forward into the design evaluation stage: some VHH-Fc candidates showed good expression levels in CHO cell expression experiments at a 10 mL culture scale; many purified candidates were dominated by a non-aggregated main peak component, with few signs of aggregation; and the tested candidate subset also showed low nonspecific reactivity against two surrogate antigens, insulin and double-stranded DNA.

While pursuing binding activity, the platform has already taken into account the key properties required for molecules to enter subsequent drug development.

Sequence novelty is even more direct evidence of "design without reference answers."

MMDesign's generation process does not rely on retrieving known binding molecules from antibody sequence databases as starting points for engineering; comparisons with databases such as PDB, OAS, and SAbDab show that designed molecules for multiple targets differ significantly from known antibodies in the antigen recognition region, and the key binding loop region CDR3 of some molecules presents novel conformations.

The candidates produced by the platform are not "copies" of known molecules, but new sequences generated around functional requirements, directly targeting problems that traditional R&D struggles to overcome, and also offering the possibility of bypassing existing patent layouts and expanding entirely new candidate space.

High-precision structure prediction at twice that of AlphaFold 3, and the "new AI biology infrastructure" behind it

For de novo design to "select accurately," structure prediction must first "see accurately."

MMFold, the all-atom structure prediction model self-developed by MoleculeMind, is the structural foundation behind MMDesign.

On the internationally authoritative FoldBench antibody-antigen complex structure prediction benchmark, MMFold achieved a Top-1 high-quality structure prediction success rate of 39.0% across all 172 interfaces evaluated, about twice the 19.8% comparison result of AlphaFold 3;

The Top-1 overall success rate reached 70.9%, the Top-5 overall success rate further reached 76.7%, among which the high-quality prediction success rate was 44.2%.

△Structure prediction results for all 172 antibody-antigen interfaces in FoldBench

On the 160 shared interface subset corresponding to the public results of OpenDDE and Protenix-v2, MMFold's Top-1 high-quality success rate reached 39.4%, higher than the 33.0% reported by OpenDDE and the 28.2% reported by Protenix-v2; the Top-5 high-quality success rate further reached 45.0%, with an overall success rate of 78.8%.

△Structure prediction results on the 160 shared interface subset of FoldBench

The value of structure prediction lies not in the ranking itself, but in giving candidate molecule evaluation and selection a more solid computational basis.

Only when structures are seen more accurately can molecular designers more quickly lock in, from the vast computational space, the few candidates worth experimenting on.

The multimodal protein foundation large model NewOrigin (Darwin) provides underlying capabilities, MMDesign undertakes de novo generation and candidate selection, MMFold focuses on structure prediction, QuantaMind performs dynamic simulation, and the AI-native R&D operating system MoleculeOS strings model tools and R&D processes into a closed loop of "generation-computation-experiment-iteration."

Xu Jinbo, founder of MoleculeMind, said:

The important significance of this expanded evaluation lies in gradually transforming one-off design successes into repeatable, scalable R&D capability. The long-term value of AI biopharmaceutical R&D is not just performance breakthroughs of individual models, but rather, through the collaboration of proprietary underlying models with generation, evaluation, and experimental validation, making precise molecular design a technical infrastructure capable of continuously solving different R&D problems.

From single-point hits to systematic validation across targets and modalities, from underlying model breakthroughs to reusable R&D workflows, MMDesign's latest report advances the question of "whether AI can design molecules" to "whether AI can stably and reproducibly produce usable molecules."

The design efficiency of few candidates with stable hits; molecular quality that balances affinity, developability, and novelty; and the breakthrough value of not relying on "reference answers," directly producing molecules from functional requirements, and successfully tackling high-difficulty targets. Together, the three support "reusability," enabling AI molecular design to be continuously used by industry to solve real R&D needs.

"Special statement: The above content (including videos, pictures, and audio if any) is uploaded and published by users of "Dafeng Hao," a self-media platform under Phoenix News. This platform only provides information storage space services.

Notice: The content above (including the videos, pictures and audios if any) is uploaded and posted by the user of Dafeng Hao, which is a social media platform and merely provides information storage space services."