Same target, four ADCs, one approval - what actually separated them?

How to build a best-in-class FRα ADC

Same target, four ADCs, one approval — what separates clinical success from development risk, and where the engineering opportunity lies.

1. Introduction

Folate receptor alpha (FRα, encoded by FOLR1) is a GPI-anchored membrane protein that mediates folate uptake by receptor-mediated endocytosis. It is overexpressed in most epithelial ovarian cancers with limited normal-tissue expression — making it one of the most studied ADC targets in gynaecological oncology. Four companies independently built an FRα ADC, each making different antibody, linker, and payload choices.

Despite the shared receptor, the agents sit at very different stages of clinical maturity. The question is whether the competitors can close the gap by learning from the design choices behind Mirvetuximab’s progress — or whether they carry molecular disadvantages that patient selection alone cannot overcome.

3.1 ADC-specific properties: binding, epitope, payload

Binding affinity. Mirvetuximab has the tightest measured binding (Kd = 1.8 nM). Farletuzumab (MORAb-202/Eisai) is weakest at 6.5 nM — a 3.6× gap. STRO-002 and Rinatabart lack published Kd; proxy estimates place them at 3.42 nM.

A 3.6× affinity gap is clinically meaningful: in moderate-FRα tumours, higher-affinity antibodies achieve greater receptor occupancy and internalise more ADC per cell. The 1.8 nM benchmark sets a competitive floor any best-in-class programme should match or beat.

Epitope positioning. All known clinical-stage anti-FRα antibodies converge on the aa45–57 region (~15.89 Å from the folate pocket), so they do not compete with folate for binding. That convergence limits epitope-based differentiation today — but a novel surface patch could differentiate a future agent, at the cost of unknown internalisation kinetics.

Epitope binding comparison. Mirvetuximab shows the highest affinity (Kd = 1.8 nM); all agents cluster at the same epitope distance (~15.89 Å), reflecting convergence on the aa45–57 immunodominant region. Triangle markers = proxy-imputed values.

Payload permeability — the most differentiating dimension. Unlike affinity and epitope, where the agents are similar, payload permeability varies strikingly. The four payloads span all three permeability classes.

Payload permeability profile (XLogP vs TPSA). Hemiasterlin (Sutro’s STRO-002) falls in the High zone; DXd (Genmab/AZ’s Rina-S) is Low.

3.2 Patient-selection strategy

Mirvetuximab’s programme is defined by biomarker-guided selection: 19 of its 26 trials (73.1%) require explicit FRα thresholds, many demanding ≥75% tumour positivity by IHC. The other agents have smaller, earlier-stage programmes with lower proportions of biomarker enrichment.

Distribution of FRα threshold classes across the four ADC programmes.

Patient selection by IHC is replicable, not proprietary — any competitor can adopt the ≥75% threshold (recent STRO-002 data already shows ORR 43.8% at FRα >25%). The distinction that matters: patient selection gets you into the game; intrinsic ADC properties determine whether you win it.

3.3 Does high FRα expression predict efficient processing?

If FOLR1 were tightly co-regulated with lysosomal proteases and endosomal trafficking, patient selection alone might guarantee efficacy. If not, there is a gap that ADC engineering — payload permeability — must fill.

FOLR1 mRNA expression across ~300 TCGA ovarian cancers. mRNA is a proxy for protein; surface abundance is shaped by post-translational processing, GPI-anchor assembly, recycling and internalisation, none captured by bulk RNA-seq.

FOLR1 vs processing markers. Only LAMP1 and CTSD show weak but statistically significant correlations.

High FOLR1 mRNA does not reliably predict abundant lysosomal enzymes or efficient trafficking. The strongest correlation (CTSD, r = 0.20) explains only ~4% of the variance; CTSB and RAB5A show essentially none. So even in IHC-selected patients, cells vary widely in how efficiently they process an internalised ADC — a biological rationale for permeable payloads that goes beyond expression heterogeneity, and one patient selection cannot replicate.

3.4 Integrated comparison

Bringing the ADC-specific and patient-selection dimensions together reveals where each agent leads and where it is exposed.

Radar chart comparing the four FRα ADCs. Mirvetuximab bulges toward Clinical (selection) and Affinity but is middling on Permeability; STRO-002 bulges toward Permeability but is weak on Clinical.

Dimension scores by agent. Mirvetuximab’s advantage is primarily programme design; STRO-002’s is primarily intrinsic ADC properties. The best-in-class question is whether one molecule can combine both.

4. What would a best-in-class FRα ADC look like?

Mirvetuximab’s lead is built on good — but not exceptional — ADC properties plus an exceptionally well-designed selection programme. On intrinsic properties alone, the picture is more open. Read straight off the data, the target design specification is:

5. Conclusion: the path to best-in-class runs through engineering

The data supports a specific reading: Mirvetuximab succeeded primarily because of its patient-selection programme, and there is meaningful room to improve on its intrinsic engineering. Patient selection is necessary to demonstrate efficacy in a regulatory context; payload and antibody design are what create a genuinely differentiated molecule. The three priorities for any FRα ADC programme: increase payload permeability, target affinity ≤ 2 nM, and treat ≥75% IHC selection as standard practice rather than a moat.

This analysis is illustrative, not definitive, and was reviewed by a human researcher for scientific accuracy. It demonstrates how an agentic AI workflow compresses weeks of manual landscape research into hours — and the same framework can be repointed at any target, indication, or competitive set.