August 2023 — Monthly Briefing¶
1. What was read¶
752 research articles from 33 of the 33 journals on the roster.
A further 197 records (21% of the 949 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 |
|---|---|---|---|---|
| 444 | 160 | 80 | 245 | 20 |
Largest contributors: Nature 124, Science 86, Nucleic Acids Res 86, JAMA 85, Bioinformatics 68, Lancet 46.
2. Net change on the thread curve¶
| Thread | This month | Cumulative | Status |
|---|---|---|---|
cell-therapy |
+4 / −1 | +57 / −18 | forming |
ctdna-mrd |
+1 | +17 / −10 | splitting |
foundation-models |
+5 | +20 / −11 | splitting |
immunotherapy-resistance |
+2 / −5 | +73 / −20 | forming |
interception |
+3 | +46 / −4 | forming |
method-credibility |
−7 | +1 / −90 | crisis |
neoantigen-vaccines |
+5 / −3 | +27 / −10 | forming |
neuro-immune-microbiome-axis |
+2 | +46 / −4 | forming |
next-gen-checkpoints |
+2 / −1 | +13 / −11 | splitting |
single-cell-3d-genome |
+2 | +15 / −5 | forming |
suppressive-population-inversion |
+1 / −3 | +17 / −18 | contested splitting → 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¶
foundation-models gains five cards and they are four different claims.
Knowledge: 37438534 (Nature)
scores a general language model against clinical question sets and reports it as
encoding clinical knowledge, with JAMA covering the same class of result
37467004. Design:
37433327 (Nature) generates
protein backbones and functions that hold up experimentally, with a language
model writing functional sequences across diverse families
36702895 (Nat Biotechnol, News &
Views 36702894) and a universal
zinc-finger design model 36702896
alongside it — three groups across two journals on the claim that these models
emit biology rather than predictions about biology, with mega-scale folding
stability data 37468638 underneath.
Deployment: 37541274 (Lancet
Oncol) compares AI-supported reading against standard double reading inside a
randomised screening programme, on a clinical safety endpoint. Transfer:
37829001 (Nat Mach Intell) trains
pan-allelic networks on mass-spectrometry ligands and transfer-learns
immunogenicity, beating seven prior models on neoepitope precision — the gain
coming from transfer between a large task and a small one, which is the argument
for foundation models stated in its own terms and on a cancer problem. The status
line says splitting; these cards are not in tension, they are in different
layers.
A microbiome intervention with a human readout, which is the thing this thread
has been short of. 37414899 (Nat
Med) gives faecal microbiota transplantation with anti-PD-1 in a phase 1
melanoma trial. Beside it 37540734
(Sci Immunol) has microbial energy metabolism sustaining a macrophage niche
inside isolated lymphoid tissue through purinergic signalling — microbial,
lymphoid-structural and small-molecule control inside one mechanism.
neuro-immune-microbiome-axis is the corpus's largest thread and its standing
weakness is that the microbial arm stays in mice; this month it is dosed in
people.
Precancer as an object that can be found, read and acted on, in one month. 37537298 (Nat Cancer) uses protein-level mutant p53 reporters to pull rare precancerous clones out of histologically normal tissue and shows them to be druggable. 37500728 (Nat Genet) reads cell-free DNA as single-molecule genome-wide mutation profiles instead of a fixed panel of known variants. And 37646678 (NEJM) puts an anti-Epstein-Barr virus antibody test into mass screening for nasopharyngeal carcinoma.
Grade: gene-edited autologous cells treat a non-malignant disease in NEJM.
37646679 edits the HBG promoters in
patients' own haematopoietic cells for sickle cell disease, with the delivery-side
work done by lipid nanoparticles in the same month
37294917 (Blood); JAMA separately
notes an FDA cell-therapy action in type 1 diabetes
37436739. cell-therapy closes
+4 / −1, and the other three supports are in patients or primates rather than in
principle: a mesothelin-targeted TCR fusion construct returns a 20% response rate
and 77% disease control in refractory solid tumours
37501016 (Nat Med); the TCRs
persisting after curative TIL therapy each recognise three unrelated tumour
antigens through one shared short motif
37490916 (Cell); and NKG2D-directed
CAR T cells clear naturally occurring senescent cells in aged mice and non-human
primates 37585504 (Sci Transl Med),
which is the engineered-cell platform pointed at something that is not a tumour at
all. The single refutation is in section 4.
Corrected 2026-09-04
This paragraph closed "cell-therapy still closes level at +1 / −1, for the
reason in section 4", and the word doing the work was level — one escape
cancelling one advance. The month is +4 / −1 and the thread reads forming.
August 2023 had been carded from titles; the 2026-09-04 abstract re-scan read
the abstracts and found three more supports, all of them clinical or primate.
The escape in section 4 is real and unchanged; it was never the whole of the
month.
4. Refutations¶
(a) Landing on a tracked thread¶
The germline strain on next-gen-checkpoints, now stated twice by two groups.
37648262 (J Immunother Cancer)
reports a germline NKG2A variant that predicts immunotherapy outcome and reshapes
the NK repertoire. Read with last month's
37264229 (Nat Immunol), where an
HLA class I signal peptide polymorphism sets how hard the same HLA-E axis presses
on effector cells, two journals one month apart have said that the magnitude of
this checkpoint's effect is a property of the host rather than of the target.
Everything the thread currently records is registry activity — whether a
nominated target reached a phase 1. This is a different and worse kind of strain,
because a target can be perfectly druggable and still deliver a different-sized
benefit to every patient, which would make both the unstratified preclinical
nomination and the unstratified trial the wrong instrument for measuring it. It
also cannot be seen from a trial registry at all, which is where this thread has
been doing its checking.
37647386 (Sci Transl Med) — activating TREM1 on myeloid cells promotes antitumour immunity. Inhibition is the strategy this target was nominated for, and it is the wrong direction.
37474835 (Nat Cancer, with its own commentary 37474834) — human solid tumours carry γδ T cells split into opposing cytotoxic and wound-healing arms. Neither depleting nor expanding the population has a defined sign.
37553183 (JITC) — tumour
cell-intrinsic STING, and not host STING, drives local and systemic antitumour
immunity after cryoablation. That lands against immunotherapy-resistance, whose
claim is that the decisive machinery sits outside the tumour cell; the same
thread gains a supporting card in the same month from obesity redirecting host
monocyte development to enhance metastasis
37166450 (JEM), and a second from
CCR7-negative cDC1 clustering with stem-like CD8 T cells in stromal niches
37451271 (Cancer Cell).
Four more refutations land on the same thread, and three of them fail in one place — the biomarker. Neither tumour mutational burden nor MSIsensor score predicts checkpoint response in MSI metastatic colorectal cancer 37269904 (Ann Oncol); no T-cell population associates with response to Y90 radioembolisation followed by nivolumab in hepatocellular carcinoma, and biomarkers established in prior immunotherapy studies do not transfer 37586766 (JITC); and across multiple cancers neither intratumoral TCR nor BCR repertoire diversity is an independent prognostic biomarker of survival 37604641 (JITC). The fourth is mechanistic: PD-1 blockade increases the self-renewal of stem-like CD8 T cells as fast as it accelerates their differentiation, leaving the progenitor pool intact 37624909 (Sci Immunol). +2 / −5 on one thread in four weeks is the honest state of it.
Corrected 2026-09-04
This closed "+1 / −1 on one thread in four weeks is the honest state of it", counting one support and one refutation. The month is +2 / −5. The three biomarker failures and the PD-1 self-renewal result were all carded by the 2026-09-04 abstract re-scan; none of them is visible in a title. The reading the sentence was defending — that a single STING result was the whole strain on this thread in August — was too small, and the strain is mostly about what can be measured before treatment rather than where the machinery sits.
37474833 (Nat Cancer) — leukaemia switches lineage after CD19-directed CAR T and blinatumomab, escaping by changing what the cell is rather than what it displays. 37607769 (JITC) — an idiotype-neoantigen vaccine raises therapeutic T cells and the heavy-chain-loss escapees cause disease. Two antigen-directed platforms, two antigen-independent escapes, one month.
(b) Method credibility¶
Seven cards. The sign is structural, as it is every month on this thread: it is a ledger of things found unreliable and can report nothing but refutations, so read what is being audited rather than the ratio.
37563719 (Genome Biol) — a suspect list of genomic regions that produce CUT&RUN signal regardless of the antibody. Peak sets published before such a list existed carry artefacts nobody filtered, because there was nothing to filter against.
37429993 (Nat Methods) — a
significance test for single-cell clusters, which implies the populations
routinely reported from scRNA-seq are not established to be distinct. With
37449901 (Bioinformatics) last
month, two independent groups in consecutive months have built tools to ask
whether the field's basic unit of analysis is real. That is the healthy reading
of crisis on this thread: not a collapse, an audit that has reached the bottom
layer.
37386187 (Nat Methods) — running single-coverage against multi-coverage metagenomic binning on the same samples exposes contaminant contigs and chimeric bins the standard approach produces silently. 37443337 (Nat Methods) — sixteen circRNA callers on the same deeply sequenced cells return between 1,372 and 58,032 calls. Which tool was used determines what the paper reports. Three of this month's method cards are in one journal.
37459870 (Lancet Oncol) — a phase 2 immunotherapy trial in melanoma with CNS metastases is retracted and republished. The numbers as originally read were wrong, in a trial already in the literature.
The last two are the JITC biomarker papers already described in (a): 37586766 and 37604641. They are carded on both threads because a biomarker that does not replicate is at once a fact about the biology and a fact about the measurement.
Corrected 2026-09-04
This section was headed "Three cards" and listed three. There are seven. The
four that were missing — two Nat Methods benchmarks, and the two JITC
biomarker papers that also land on immunotherapy-resistance — were carded by
the 2026-09-04 abstract re-scan. The structural point about the sign is
unaffected; the audit was three times the size stated.
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 | 631 (of which 552 genuinely oncology) |
| fda-approval | 6 |
| ema-authorisation | 3 |
6 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:
- 2023-08-02 — FDA approves trifluridine and tipiracil with bevacizumab for previously treated metastatic colorectal cancer
- 2023-08-09 — FDA grants accelerated approval to talquetamab-tgvs for relapsed or refractory multiple myeloma
- 2023-08-09 — FDA approves pralsetinib for non-small cell lung cancer with RET gene fusions
- 2023-08-11 — FDA approves niraparib and abiraterone acetate plus prednisone for BRCA-mutated metastatic castration-resistant prostate cancer
- 2023-08-14 — FDA approves melphalan as a liver-directed treatment for uveal melanoma
- 2023-08-14 — FDA grants accelerated approval to elranatamab-bcmm for multiple myeloma
1 EMA oncology authorisation:
- 2023-08-21 — Talvey
8. Trade press¶
No items. The press archive covers 2024-01 to 2026-08; feeds that cannot be paginated backwards contribute only from the first harvest forward.