August 2025 — Monthly Briefing¶
1. What was read¶
750 research articles from 32 of the 33 journals on the roster.
A further 271 records (27% of the 1,021 this month's harvest returned) carry no abstract and are not research: news, correspondence, editor highlights and obituaries. PubMed types them Journal Article and nothing else, so the publication-type filter cannot see them — across the whole index they are 22.7% of it and 35% of general, and 3,801 of them are JAMA. Every count below is over the full harvest and is inflated by this; the figure above is not.
| General | Cancer | Immune | Bioinfo | Sysbio |
|---|---|---|---|---|
| 490 | 166 | 93 | 242 | 30 |
Volume swing against the median month: sysbio +25%
Check the source before reading anything into this. Two benign causes account for most of these. bioinfo every January carries the Nucleic Acids Research database issue and runs high. And several journals simply deposit nothing in some months: across the 44 months read, 4 of 33 journals are absent from at least one, worst being Nat Mach Intell (20 months missing, 81 papers in total) and Bioinformatics (4 months missing, 2,905 papers in total). This month is missing Nat Mach Intell. Queried directly, PubMed returns zero for those journal-months too — the holes are in the source, reproduced exactly by re-running the harvest. So a swing here is not evidence that a field went quiet, and is not usually evidence that the pipeline broke either.
Largest contributors: JAMA 136, Nature 103, Nucleic Acids Res 86, Science 80, Lancet 45, Bioinformatics 42.
2. Net change on the thread curve¶
| Thread | This month | Cumulative | Status |
|---|---|---|---|
agentic-ai |
+2 | +14 / −1 | forming |
cell-therapy |
+2 / −2 | +189 / −60 | forming |
ctdna-mrd |
+4 / −1 | +69 / −23 | forming |
foundation-models |
+6 / −4 | +108 / −76 | splitting |
immunotherapy-resistance |
+3 / −1 | +210 / −75 | forming |
interception |
+1 / −1 | +118 / −23 | forming |
method-credibility |
−14 | +18 / −429 | crisis |
neoantigen-vaccines |
−3 | +76 / −36 | forming |
neuro-immune-microbiome-axis |
+6 | +163 / −17 | forming |
next-gen-checkpoints |
−1 | +66 / −39 | splitting |
single-cell-3d-genome |
— | +65 / −17 | forming |
suppressive-population-inversion |
+1 / −2 | +50 / −77 | contested |
Status is derived by rule, not asserted. Fewer than 5 cards → too-early; then refutes ≥ 2×supports → crisis; refutes > supports → contested; supports ≥ 2×refutes → forming; else splitting.
3. The evidence that moved it¶
Clinical AI acquired trial-grade evidence on the enthusiastic side. An eyecare foundation model was published as a randomized controlled trial and a generative-AI-discovered TNIK inhibitor reached a randomized phase 2a, while the skeptical side of the month is two benchmarks of open-weight models. The caveat is load-bearing and unresolved: a randomized AI trial is frequently randomized on workflow or efficiency endpoints rather than on patient outcomes, and neither abstract has been read.
Corrected 2026-08-26
This section originally called August an inversion — the claim being that through spring the enthusiastic side had benchmarks while the skeptical side had randomized trials, and that August reversed it. The second half of that was never true. Re-reading January to July 2025 under one consistent rule found zero titles stating a null or negative result for any AI intervention in any month. Randomized AI trials existed from February (GPT-4 assistance, TAILORED-AF) and none of them reported a null in its title. What spring actually had was a methodological critique — A benchmarking crisis in biomedical machine learning, April 2025 — plus a single-cell foundation-model limitations paper. Nothing inverted in August; the enthusiastic side gained, and the skeptical side was never where this corpus had recorded it.
Microbiota-to-antitumour-immunity got a cell, not a taxon. 40659786 attributes microbiota-driven antitumour immunity to dendritic cell migration. Most of what this corpus has recorded on the microbiome-immunotherapy link is compositional correlation; this is a trafficking mechanism, and it is the kind of result that decides whether the thread survives contact with intervention trials.
Cancer neuroscience was the crowded direction of the month — three Nature
papers and two Cell papers, including
nerve injury driving anti-PD-1 resistance,
which is the first time in this corpus the neural axis is joined to checkpoint
resistance. The reviewing agent flagged the cluster as peaking on the grounds
that it is entirely preclinical with zero clinical readouts. That call has now
been made three times on this thread and has been wrong three times. See the
note on neuro-immune-microbiome-axis in data/threads.yml: a direction with
negative clinical readouts is failing, a direction with no clinical readouts is
untested, and the literature alone cannot tell them apart.
4. Refutations¶
(a) Landing on a tracked thread¶
Fourteen cards landed refutes on a tracked thread other than
method-credibility this month, across eight of them; two of the fourteen also
carry method-credibility, including the connectomics limit below.
The three that bite hardest each attack a premise rather than a product. 40759747 (Nat Methods) states it in its title: deep-learning gene-perturbation-effect prediction does not yet outperform simple linear baselines — the task perturbation foundation models were built for. 40425843 (Nat Med) is a prospective multi-cancer early-detection evaluation of cell-free DNA fragmentomics returning 53.5% sensitivity, well below what the interception programme assumes of a screening assay. 40405018 (Nat Cancer) reports that tumour antigens in melanoma and NSCLC derive preferentially from unmutated genomic sequence, which is the mutated-neoantigen premise itself rather than any one vaccine.
Cell therapy took two clinical readouts in the same direction: intraventricular bivalent EGFR/IL-13Rα2 CAR-T in recurrent glioblastoma reached a median progression-free survival of 1.9 months (40451950), and one-time autologous TIL in recurrent or metastatic head and neck squamous cell carcinoma returned an 11% objective response rate (40854613). Neither is a delivery-route or manufacturing failure; both are the solid-tumour extension not extending.
The month's largest negative block outside the tracked threads — three simultaneous aspirin-withdrawal trials, two beta-blocker-after-MI trials, and antihypertensive deprescribing — sits in cardiology, outside the corpus fields, and is recorded here only because it is the clearest de-escalation programme in the eighteen months scanned.
Corrected 2026-09-04
This subsection read "Nothing landed on a tracked thread with a title-stated direction this month beyond the connectomics limit below", and had nothing in it but the cardiology block. That was true of the titles and false of the month. August 2025 was carded from titles alone, and the qualifier "with a title-stated direction" was doing the work of a denominator: a clinical readout that reports its number in the abstract and not in the title was invisible to the reader, so an empty subsection recorded the reading method rather than the literature. The 2026-09-04 abstract re-scan finds fourteen refutations on tracked threads here, including two phase-1/2 efficacy readouts and a title-stated null on foundation models. The qualifier is retired; it cannot mean anything in a corpus that reads abstracts.
(b) Method credibility¶
Two Nature retractions in one month, both of anchors rather than minor papers: the Parkin/PINK1 restraint on STING-induced inflammation and the fetal cerebellar atlas that assigned medulloblastoma its cell of origin. A retracted cell-of-origin atlas is worse than a retracted mechanism, because atlases are cited as reference rather than as claim.
40460825 (Cell) states a boundary
on the connectomics premise: structure-to-function prediction fails where
infrequent strong synapses dominate. It is filed under foundation-models as
well as method-credibility, because the assumption it limits — that a
sufficiently complete structural map predicts function — is the same assumption
under scaled models of biology generally.
40866696 (Nature) refutes the niche-limitation assumption in haematopoietic stem cell biology in its title.
5. Open items¶
Nothing carried forward.
6. Meetings¶
No abstract book published this month. The abstract layer runs 2023-02 to 2026-08 and is seasonal by nature — meetings are annual, so most months are empty here and that is the expected shape, not a gap in the harvest.
7. Regulatory¶
| Event | Count |
|---|---|
| trial-first-posted | 615 (of which 535 genuinely oncology) |
| ema-authorisation | 13 |
| fda-approval | 2 |
2 FDA actions — dated from each notice's own prose, not the table's posting date, which disagrees on 7 of 190 rows and once by enough to move an approval into the wrong month:
- 2025-08-06 — FDA grants accelerated approval to dordaviprone for diffuse midline glioma
- 2025-08-08 — FDA grants accelerated approval to zongertinib for non-squamous NSCLC with HER2 TKD activating mutations
4 EMA oncology authorisations:
- 2025-08-14 — Ogsiveo
- 2025-08-18 — Vivlipeg
- 2025-08-22 — Nintedanib Viatris
- 2025-08-25 — Zemcelpro
8. Trade press¶
30 on-topic of 282 items across 3 feeds (n-stat 272, n-aacr-blog 6, n-acir 4). Headlines are selected by subject keyword, not recency — most of what these feeds carry is health policy and industry finance the corpus does not read.
- 2025-08-28 — Opinion: STAT+: Why Silicon Valley should demand clinical trials for its medical AI —
n-stat - 2025-08-27 — Editors’ Picks, August 2025: Liquid Biopsy for Brain Cancer, Therapeutic Strategies for Prostate Cancer, and More —
n-aacr-blog - 2025-08-27 — STAT+: Pharmalittle: We’re reading about a Massachusetts biotech slump, Akeso lung cancer survival data, and more —
n-stat - 2025-08-27 — A major childhood brain cancer network is ending —
n-stat - 2025-08-27 — Overcoming dysfunctional CAR-T in pancreatic cancer —
n-acir - 2025-08-26 — STAT+: Akeso says its closely watched drug improved patient survival in a lung cancer trial —
n-stat - 2025-08-26 — STAT+: Eli Lilly’s obesity pill helped lower weight, blood sugar in key Phase 3 trial —
n-stat - 2025-08-21 — STAT+: FDA approves Ionis drug that prevents swelling attacks caused by a rare disease —
n-stat - 2025-08-21 — STAT+: FDA ‘uncertainty’ forces Krystal Biotech to shutter cancer drug trial —
n-stat - 2025-08-20 — 2025 AACR Journal Awards Highlight Broad Scope of Oncology-advancing Research —
n-aacr-blog - 2025-08-20 — STAT+: GLP-1s, cancer care are driving higher employer health care costs in 2026 —
n-stat - 2025-08-20 — Agonist anti-CD40 shows promise in a phase 1 clinical trial —
n-acir - … and 18 more on-topic items.