Weekly Intelligence Digest — 28 September–4 October 2026

Microsoft Research introduces Quine, a multimodal biology system that closes the loop between AI prioritization and wet-lab experiments

Quine combines a multimodal biological world model with reasoning, literature and scientific tools. In pancreatic cancer work with Broad Institute researchers, Microsoft reports that Quine ranked compounds predicted to shift tumor cell states and that several top-ranked candidates were validated across multiple wet-lab assays, including an unexpected third phenotype.

Why it matters: This is more interesting than another benchmark because the computational hypotheses were experimentally challenged. The evidence is still company-reported, early-stage and research-only; no clinical efficacy or independent replication has been established.

AMT-130's 48-month cUHDRS result loses statistical significance, putting external-control robustness under the microscope

uniQure reported that 12 high-dose patients had 44% slower cUHDRS decline versus an updated ENROLL-HD external control at 48 months, but the prespecified primary endpoint was not statistically significant (p=0.144). TFC showed 61% slower decline (nominal p=0.008). At 36 months, 15 high-dose patients showed 80% slowing on cUHDRS and 67% on TFC, both nominally significant.

Why it matters: The audit issue is the comparator: missingness in the matched external control reached 53% at 48 months, and post hoc use of the prior control materially changed the effect estimate. The new 48-month data were not part of the submitted BLA.

Lantheus doses the first patient in the Phase 2/3 SPARROW trial of GRPR-targeted PET in prostate cancer

SPARROW is a registrational study of 68Ga-DOTA-RM2, a GRPR-targeted PET agent with a biologically distinct target from PSMA. Phase 2 will optimize activity and imaging time while assessing biodistribution and dosimetry; Phase 3 is expected to enroll about 400 participants, with sensitivity and specificity versus histopathology as co-primary endpoints and comparative summaries versus conventional imaging and/or PSMA PET where applicable.

Why it matters: This expands the prostate theranostic target landscape beyond PSMA, but it is a diagnostic-development milestone, not evidence of therapeutic benefit.

Sun Pharma takes ex-US/China rights to once-monthly PCSK9 inhibitor lerodalcibep

LIB Therapeutics and Sun Pharma announced an exclusive global licensing, manufacturing and commercialization agreement outside the United States and China on 28 September. Lerodalcibep is already FDA-approved and received EU approval on 21 September; its Phase 3 LIBerate program enrolled more than 2,900 patients and reported sustained LDL-C reductions of at least 60% in high-risk cardiovascular populations.

Why it matters: The competitive signal for inclisiran is now broader than efficacy: once-monthly dosing, small-volume injection, six-month ambient storage and global manufacturing/access strategy will influence real-world positioning.

FDA approves tavapadon, the first selective D1/D5 dopamine-receptor agonist for Parkinson's disease

JUVMO (tavapadon) was approved on 28 September for adults with Parkinson's disease. In the Phase 3 TEMPO program it improved activities of daily living and motor scores in early disease and, when added to levodopa, increased daily "on" time without troublesome dyskinesia by 1.7 hours versus 0.6 hours with placebo at Week 26.

Why it matters: The approval adds a mechanistically distinct dopaminergic option after decades dominated by levodopa and D2/D3 agonists. The evidence supports symptomatic control; it should not be interpreted as disease modification.

OpenADMET sharpens open-science governance while its blinded CYP-inhibition benchmark remains active

OpenADMET published updated open-science rules on 29 September while running its CYP inhibition blind challenge across CYP3A4, CYP2C9, CYP2D6 and CYP1A2. The project provides open models, curated experimental datasets and blinded test sets rather than relying on retrospective leaderboard performance.

Why it matters: That makes it unusually useful for drug-discovery AI: models can be tested against withheld experimental data relevant to drug-drug-interaction risk. It remains research infrastructure, not a validated regulatory substitute for experimental ADMET testing.

FDA Rare Disease Innovation Hub / Duke-Margolis Institute for Health Policy

The 29 September RISE workshop focused on randomization, blinding, efficiency and tailored statistical approaches in small rare-disease populations.

Why it matters: The useful regulatory signal is that flexibility is being discussed as a way to preserve interpretable evidence under severe sample-size constraints, not as permission to relax statistical rigor.

FDA CDER and CBER

On 1 October FDA began accepting year-five applications to the Chemistry, Manufacturing, and Controls Development and Readiness Pilot. Selected accelerated programs receive additional Type B interactions focused on CMC strategy.

Why it matters: For fast clinical programs, manufacturing readiness is being treated explicitly as a development-rate limiter rather than paperwork to solve near filing.

FDA cleared PM647, an in vivo Prime Editing therapy for alpha-1 antitrypsin deficiency, to enter clinical testing on 24 September 2026

Why still relevant: The one-time liver-directed editor is designed to correct the Pi*Z SERPINA1 mutation and address both pulmonary deficiency and hepatic toxic-protein accumulation. Initial clinical data are expected in 2027.

FDA granted final approval to BRAVNETSA on 22 September 2026

Why still relevant: BRAVNETSA is the first lutetium Lu 177 dotatate product FDA has determined to be bioequivalent and therapeutically equivalent to Lutathera, adding another supply option for SSTR-positive GEP-NET radioligand therapy.

FDA approved Fayuvi, the first gene therapy for pediatric Sanfilippo syndrome type A, on 17 September 2026

Why still relevant: The approval relied on an open-label single-arm study compared with an untreated natural-history cohort, making it an important recent example of external-control evidence supporting full approval in an ultra-rare disease.

FDA launched the Expedited IND Pilot on 15 September 2026

Why still relevant: Selected sponsor-research-institution pairs can use rolling review of IND components during the pre-IND phase, potentially moving issue resolution, IRB planning and site activation earlier in first-in-human development.

Phase 3 Lp(a)HORIZON missed its cardiovascular-outcomes endpoint despite lowering Lp(a), reported 4 September 2026

Why still relevant: Pelacarsen's failure remains the strongest recent warning against treating profound biomarker lowering as proof of cardiovascular benefit and raises the evidentiary bar for the wider Lp(a) field.

AI-assisted discovery: require wet-lab confirmation before prioritization

For AI-supported discovery, require traceable hypothesis generation followed by blinded or wet-lab confirmation before AI output influences candidate prioritization.

Gene therapy: pre-specify external-control attrition and sensitivity rules

For gene-therapy programs using external controls, pre-specify attrition handling, missing-data sensitivity analyses and comparator refresh rules; AMT-130 shows how longitudinal missingness can materially alter apparent treatment effect.

Prostate theranostics: design truth standards against established PSMA imaging

For new prostate radiopharmaceutical targets, design truth standards, reader reproducibility, dosimetry and comparator analyses against established PSMA imaging before pivotal expansion.

Lipid strategy: model dosing, storage, and access alongside LDL-C reduction

In lipid competitive strategy, model route, dosing frequency, storage, manufacturing and access alongside LDL-C reduction; differentiation is moving beyond pharmacodynamic potency alone.

Accelerated programs: bring CMC risk onto the critical path early

For accelerated clinical programs, bring CMC risk onto the critical path early and use additional regulator interaction where eligible; clinical speed is useless if process validation and manufacturing readiness lag behind.