Skip to content

July 2024 — Monthly Briefing

1. What was read

842 research articles from 33 of the 33 journals on the roster.

A further 110 records (12% of the 952 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
358 165 81 328 20

Largest contributors: Nucleic Acids Res 162, Nature 99, Science 80, Bioinformatics 62, J Immunother Cancer 50, JAMA 42.

2. Net change on the thread curve

Thread This month Cumulative Status
agentic-ai — +4 / −0 too-early
cell-therapy +7 / −2 +115 / −38 forming
ctdna-mrd +3 / −2 +37 / −19 splitting forming → splitting
foundation-models +4 / −4 +52 / −37 splitting
immunotherapy-resistance +10 / −12 +132 / −51 forming
interception −2 +69 / −15 forming
method-credibility −22 +6 / −252 crisis
neoantigen-vaccines +1 / −3 +50 / −22 forming
neuro-immune-microbiome-axis +3 / −2 +93 / −13 forming
next-gen-checkpoints +3 / −2 +38 / −29 splitting
single-cell-3d-genome +3 +38 / −15 forming
suppressive-population-inversion +5 / −1 +32 / −41 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

A spatial unit, not a cell type, is what immunotherapy requires. 38906155 (Cancer Cell): intratumoral immune triads are required for immunotherapy-mediated elimination of solid tumours. The operative requirement is a multicellular arrangement. Read against the whole suppressive-population-inversion thread, this is the constructive version of the same objection: if the functional unit is a triad, then depleting or blocking any single population in it has an unpredictable sign, which is exactly what the inversion cards keep reporting.

Resistance that costs the cell its tumour. 38533987 (Cancer Discov) inhibits protein phosphatase 2A in colon cancer cells, which hyperactivates several oncogenic pathways at once and engages the stress responses that aberrant signalling provokes. Genetic and compound screens find combined PP2A and WEE1 inhibition synergistic across models: replication collapses, mitosis comes early, the cells die, and patient-derived tumours shrink in vivo. The part worth keeping is the resistance phenotype — cells that acquire resistance to the combination lose oncogenic signalling and with it the ability to form tumours at all. Tumour-suppressive drug resistance is a genuinely unusual object and it belongs to no thread here.

Corrected 2026-08-27

Three errors came out of this one title. The entry read: "The sign of an intervention is treated as an open question in a review-of-record venue — a year before the corpus starts logging sign reversals as a pattern." The record is not a commentary but a primary research article with genetic and compound screens and PDX work, so there is no review-of-record venue treating anything as open. It was carded on next-gen-checkpoints, and contains no checkpoint, no immune target and no antibody. And the gloss about the sign of an intervention has no counterpart in the paper. The card is now neutral, typed mechanism, and carries no thread.

4. Refutations

(a) Landing on a tracked thread

Twenty-five of this month's refutations land on a tracked thread other than method-credibility, and they fall into three groups.

Five randomised trials read out negative, across four threads. 38755096 (Ann Oncol) — IMpassion132 is a fully negative double-blind phase III of atezolizumab added to chemotherapy in early-relapsing triple-negative breast cancer, with no overall survival gain in any PD-L1 stratum. 38552197 (JCO) — CONTACT-01 is a negative phase III of atezolizumab plus cabozantinib after checkpoint failure, so re-challenging with an antiangiogenic does not overcome acquired resistance. 38704093 (Ann Oncol) — in MITO END-3, avelumab helped only the p53-abnormal subgroup and not the cohort, so molecular profile rather than checkpoint biology decided who benefited. 38588588 (Cancer Discov) — a randomised placebo-controlled phase Ib of the microbiome consortium SER-401 with antibiotic preconditioning produced a lower response rate than placebo, so deliberate microbiome modulation did not translate into better checkpoint outcomes — a controlled test of the intervention arm of neuro-immune-microbiome-axis, against the thread. And 38848742 (Lancet Oncol) — the randomised DENIM trial of dendritic cells loaded with allogeneic tumour lysate in mesothelioma did not improve survival over best supportive care.

Three cards put the decisive machinery back inside the tumour cell. 38869480 (JEM) — PD-L1 acts intracellularly to shift metabolism and block type I interferon signalling, a cell-intrinsic function of the checkpoint itself. 38991060 (Science) — mutant IDH1 silences cGAS in the tumour cell, and inhibiting it restores double-stranded DNA sensing and tumour immunity. 39089738 (JITC) — blocking G6PD in the tumour cell induces immunogenic cell death and accelerates immunotherapy. immunotherapy-resistance holds that the decisive machinery sits outside the tumour cell; this is the same objection August 2023 recorded once, arriving three times in Science, JEM and JITC in four weeks. The thread closes +10 / −12.

foundation-models takes four refutations and they are four different failures. 39048682 (Nature) — models trained on recursively generated data collapse, so scale cannot be extended indefinitely on model-produced corpora. 38876123 (Lancet Oncol) — in the international PI-CAI reader study the AI system was not non-inferior to radiologists for clinically significant prostate cancer on MRI, so it did not transfer into the screening role it was built for; it lands on interception too. 38913850 (Bioinformatics) — a dictionary-based concept recogniser matches or beats large language models on human-phenotype extraction while running orders of magnitude faster. 38917327 (NAR) — a comparative assessment including AlphaFold3 finds RNA 3D structure methods still model RNA–ligand interactions poorly. Data, deployment, cost-effectiveness and transfer to a second molecule, in one month.

Corrected 2026-09-04

This section read, in full: "Nothing lands on a tracked thread this month." Twenty-five refutations do, including five randomised trials and a Nature paper on model collapse. July 2024 had been carded from titles, and the trial titles here name the drug and the design and not the result — "IMpassion132 double-blind randomised phase III trial of chemotherapy with or without atezolizumab", "CONTACT-01: A Randomized Phase III Trial" — so the month's negatives were invisible to the instrument that read it. The 2026-08-27 correction below was right to remove the one card this section did carry — but removing it left the section reading as a finding about July 2024 when it was a finding about how July 2024 had been read.

Corrected 2026-08-27

This section carried 38969522 (JITC): "infliximab in steroid-refractory immune-related hepatitis does not demonstrate hepatotoxicity and may shorten time on steroids — a feared toxicity that does not appear." The finding is real — a ten-patient single-centre retrospective series in which infliximab controlled grade ≥3 immune-related hepatitis in 7 of 10 without liver injury, resolved in 9 of 10 at a median 487 days — but it is the management of an immune-related adverse event, not a mechanism of checkpoint resistance, and it was carded supports on immunotherapy-resistance. None of the other thirty cards on that thread is about toxicity. The card keeps the finding, is now neutral, and has no thread in this corpus.

(b) Method credibility

Three concern-and-retraction notices in one month, but two fewer signals than that count suggests. 38946327 and 38946326 (Cancer Discov) are Expressions of Concern on drug-addiction-based selection against MAPKi-resistant melanoma and on durable suppression of acquired MEK-inhibitor resistance. The papers are distinct and both notices stay on the ledger, but they share their first author and come out of one laboratory's body of work, so they are one investigation reaching two papers rather than two investigations. 38870944 (Cell, retraction of a genomic decoding of neuronal depolarisation) is separate. Both Cancer Discovery notices attach to resistance-mechanism papers, which is the literature the immunotherapy-resistance thread draws on.

Corrected 2026-08-27

The old phrasing — "two of them in the same journal", "the second concern notice in the same journal and month" — presented the pair as an independent second signal. Same journal, same month, therefore separate sources is the inference that also failed on the two pathology foundation models in March, in the other direction.

39030569 (Genome Biol) benchmarks methods for correcting biases of established and unknown origin in CRISPR-Cas9 screening data — the unknown half is the admission.

Corrected 2026-08-27

38951922 was in this paragraph, described as evaluating variant-effect predictors "by their ability to infer human traits rather than against curated labels, which is the same move as the 2026 sequence-to-function evaluations against real inter-individual variation". The method is as described — 24 predictors scored against rare-variant trait associations in UK Biobank and All of Us, which sidesteps the circularity of testing against training labels — but nothing in it is found unreliable. AlphaMissense outperforms the rest in both cohorts and the two rankings agree. A benchmark with no adverse finding is not an entry in a ledger of things found unreliable, and this corpus already cards that genre neutral elsewhere. The stance is now neutral; the reading forward to 2026 was hindsight and is withdrawn with it.

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 596 (of which 544 genuinely oncology)
ema-authorisation 9
fda-approval 1

1 FDA action — 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:

  • 2024-07-30 — FDA approves daratumumab and hyaluronidase-fihj with bortezomib, lenalidomide, and dexamethasone for multiple myeloma

7 EMA oncology authorisations:

  • 2024-07-24 — Cejemly
  • 2024-07-24 — Pomalidomide Zentiva
  • 2024-07-24 — Pomalidomide Krka
  • 2024-07-24 — Apexelsin
  • 2024-07-26 — Pomalidomide Accord
  • 2024-07-26 — Dasatinib Accord Healthcare
  • 2024-07-26 — Avzivi

8. Trade press

34 on-topic of 289 items across 3 feeds (n-stat 275, n-aacr-blog 9, n-acir 5). 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.

  • 2024-07-31 — From the Bench, July 2024: Fighting Chemo Brain With Light, Controlling Cancer Cell Death, and More — n-aacr-blog
  • 2024-07-31 — TAMing immunotherapy resistance by targeting TIM3+VISTA+ macrophages — n-acir
  • 2024-07-30 — Self-screening for Cancer: Home Sweet Home — n-aacr-blog
  • 2024-07-30 — STAT+: Pharmalittle: We’re reading about the U.K. nixing coverage for a cancer drug, sharing bird flu vaccine data, and more — n-stat
  • 2024-07-29 — STAT+: In a big step for liquid biopsy, FDA approves a blood screening test for colorectal cancer — n-stat
  • 2024-07-26 — Old Dogs Can Teach Cancer Researchers New Tricks — n-aacr-blog
  • 2024-07-26 — STAT+: Pharmalittle: We’re reading about a thumbs down for an Alzheimer’s drug, Pfizer gene therapy struggles, and more — n-stat
  • 2024-07-26 — STAT+: How Pfizer’s grand gene therapy ambitions crumbled — n-stat
  • 2024-07-26 — STAT+: Kamala Harris’ views on health care were shaped by her mother, cancer researcher Shyamala Gopalan — n-stat
  • 2024-07-25 — Breast cancer study reveals a paradox of mastectomy — n-stat
  • 2024-07-24 — Editors’ Picks, July 2024: Risks of CAR T-cell Therapy, Sarcoma Advances, and More — n-aacr-blog
  • 2024-07-24 — STAT+: Jiankui He, creator of CRISPR-edited children, relocates to a Chinese medical tourism hub — n-stat
  • … and 22 more on-topic items.