February 2025 — Monthly Briefing¶
1. What was read¶
778 research articles from 32 of the 33 journals on the roster.
A further 274 records (26% of the 1,052 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 |
|---|---|---|---|---|
| 471 | 186 | 83 | 283 | 29 |
Largest contributors: Nucleic Acids Res 136, JAMA 136, Nature 107, Science 57, Lancet 48, J Immunother Cancer 46.
2. Net change on the thread curve¶
| Thread | This month | Cumulative | Status |
|---|---|---|---|
agentic-ai |
+1 / −1 | +9 / −1 | forming |
cell-therapy |
+1 / −2 | +153 / −47 | forming |
ctdna-mrd |
— | +48 / −22 | forming |
foundation-models |
+4 / −2 | +84 / −56 | splitting |
immunotherapy-resistance |
+8 / −2 | +169 / −63 | forming |
interception |
+4 / −1 | +91 / −20 | forming |
method-credibility |
+1 / −15 | +12 / −342 | crisis |
neoantigen-vaccines |
+1 / −1 | +58 / −25 | forming |
neuro-immune-microbiome-axis |
+8 | +122 / −14 | forming |
next-gen-checkpoints |
+1 / −1 | +52 / −34 | splitting |
single-cell-3d-genome |
+3 | +51 / −17 | forming |
suppressive-population-inversion |
−4 | +37 / −55 | 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¶
Two agent papers in one month, and neither of them is a result. ESCARGOT is an LLM agent reasoning over biomedical knowledge graphs, and a Lancet comment argues for LLM agents as clinical teammates rather than tools. The comment is advocacy: it carries the vocabulary and nothing else, and the vocabulary is what it is carded for.
Corrected 2026-08-27
This paragraph said "the agentic thread opens, and the corpus can date it to the day", and quoted a four-day interval between the two papers above. The whole diffusion argument in the 2025-06 and 2025-12 briefings was measured from that interval.
It was voided on 2026-08-26 by the 2024 backfill, which found an LLM agent
with tools in a tracked journal in
2024-01, and the correction was
written into the 2024-01 briefing but never propagated back here. That date
has since moved again: with the reading floor at 2023-01, data/threads.yml
records autonomous chemical research with large language models
in Nature in 2023-12, with a Science paper
three weeks after it.
So no interval is quoted here now. ESCARGOT is not where the thread opens; it is where the corpus started looking, and any span measured forward from it was measuring the reading window rather than the field. "Dated to the day" was the most confident phrasing in the file and it was the least supported.
Two randomized trials of AI interventions, in the same journal, in the same month: GPT-4 assistance for physician performance and AI-individualized treatment of persistent atrial fibrillation. Neither title states a direction. This matters for a claim the corpus made and later withdrew: randomized AI trials were never rare in 2025, and none of them reported a null in its title.
A protein language model claims to simulate 500 million years of evolution (39818825, Science), and a cell atlas foundation model makes scalable search over human cells the deliverable. Both are capability claims with no evaluation adversary in the month — the audit genre is running at roughly one paper in seventy here, against one in ten by December.
4. Refutations¶
(a) Landing on a tracked thread¶
Three Treg inversions in one month, all outside oncology and all saying the same thing: meningeal Tregs preserve brain homeostasis, recirculating Tregs mediate thymic regeneration, and terminal effector Treg persistence is essential to prevent intestinal inflammation. The last is the sharpest for anyone designing a depletion strategy: the effector-Treg compartment that tumour immunology wants removed is the compartment gut immunology finds indispensable.
39789375 (Nat Immunol): low-avidity T cells drive endogenous tumour immunity in mice and humans. The selection logic under neoantigen prioritisation ranks candidates by predicted binding strength; this says the cells that matter are the ones that logic discards.
39972134: clonal driver neoantigens are lost under EGFR TKI and immune selection pressure — the first of what becomes a recurring attack on the size of the target space.
(b) Method credibility¶
39870925 (Nature): retraction of a published mechanism for the suppression of homologous recombination in G1 cells.
39820264 (JAMA): limited evidence supports more than forty predictive machine learning algorithms already in use in primary care. The gap between deployment and evidence is stated as a count.
39660643: across seven histone
marks, eleven cell types and a range of cell states, no single mark is the
strongest predictor of expression, and the published in silico models had omitted
mark function, regulatory distance and cell state. The audited models are
convolutional and attention networks the authors trained for the task, and the
paper ends by endorsing their use to uncover new biology — so this is a critique
of what earlier work left out of the design, not of the method class. An earlier
version of this entry filed it against foundation-models as well and read it as
capability turning out to be a property of the evaluation; that thread is off the
card and the gloss with it.
39809900: a failed reproduction, in Nature Genetics, of a published disease-resistance gene in wheat.
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 | 532 (of which 457 genuinely oncology) |
| ema-authorisation | 19 |
| fda-approval | 3 |
3 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:
- 2025-02-11 — FDA approves brentuximab vedotin with lenalidomide and rituximab for relapsed or refractory large B-cell lymphoma
- 2025-02-11 — FDA approves mirdametinib for adult and pediatric patients with neurofibromatosis type 1 who have symptomatic plexiform neurofibromas not amenable to complete resection
- 2025-02-14 — FDA approves vimseltinib for symptomatic tenosynovial giant cell tumor
6 EMA oncology authorisations:
- 2025-02-03 — Hetronifly
- 2025-02-10 — Korjuny
- 2025-02-12 — Zefylti
- 2025-02-12 — Welireg
- 2025-02-12 — Xbryk
- 2025-02-14 — Osenvelt
8. Trade press¶
25 on-topic of 299 items across 3 feeds (n-stat 286, n-aacr-blog 9, 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.
- 2025-02-27 — Unleashing the Power of CAR T Cells: Insights from Day 3 of AACR IO —
n-aacr-blog - 2025-02-27 — CDC report adds to evidence that HPV vaccine is preventing cervical cancer in U.S. women —
n-stat - 2025-02-26 — TLR agonism and PD-1 blockade unfortunately activate suppressor cells —
n-acir - 2025-02-25 — Destigmatizing Cancer in African American Communities: The Power of Open Communication —
n-aacr-blog - 2025-02-25 — Inaugural AACR IO Conference Kicks Off With Celebration of Immune Checkpoints —
n-aacr-blog - 2025-02-24 — STAT+: FDA approves first drug for rare lipid storage disease —
n-stat - 2025-02-21 — Editors’ Picks, February 2025: Improving CAR T, Studying Breast Cancer Risk, and More —
n-aacr-blog - 2025-02-21 — STAT+: Pharmalittle: We’re reading about RFK Jr. targeting CDC and FDA advisers, a Pfizer gene therapy, and more —
n-stat - 2025-02-19 — Neoantigen vaccination shows promising prevention of recurrence of high-risk kidney cancer —
n-acir - 2025-02-18 — STAT+: Duchenne gene therapy from Solid Biosciences produces promising results in early study —
n-stat - 2025-02-16 — Opinion: The abrupt end of USAID-funded clinical trials is profoundly unethical —
n-stat - 2025-02-14 — STAT+: Corbus reports first Western data for closely tracked ADC cancer drug —
n-stat - … and 13 more on-topic items.