AML drug-repurposing research is spread across many drugs, not concentrated in a few

Question

Is the published preclinical evidence for repurposing approved non-oncology drugs in AML concentrated in a few compounds?

Hypothesis · pre-registered

The top 5 candidates account for at least 50% of AML-repurposing papers in Europe PMC, 2015-2025.

Test

Harvested Europe PMC records with a script, matched titles and abstracts against a frozen 125-compound dictionary of approved non-oncology drugs, and computed the top-5 paper share (S1). The top 5 was re-selected inside each of 2,000 bootstrap resamples, and the analysis was repeated across 2 corpora × 5 definitional sensitivity scenarios.

Finding

In the primary strict corpus (Europe PMC 2015–2025, AML plus a repurposing term; 147 papers, 33 matched to a dictionary drug), the top 5 compounds account for 42.4% of drug-assigned papers. The bootstrap 95% CI is 0.39–0.67, so the primary result is below the 50% threshold but not decisive on its own. The larger loose corpus (237 papers, 159 assigned) gives 40.9% (CI 0.35–0.49), which excludes 50%. Nine of the ten scenarios fall below 0.50. The one at or above it (loose + ATRA, 0.505) only gets there by counting ATRA, an approved oncology drug. The top compound covers only 10–12% of assigned papers, and the effective number of compounds is about 19–22, which describes a long tail.

Metric	Value	95% CI	n
S1 top-5 share, strict primary	0.4242 proportion of drug-assigned papers	0.3939 – 0.6667	33
S1 top-5 share, loose primary	0.4088 proportion of drug-assigned papers	0.3522 – 0.4906	159
S1 loose + ATRA	0.505 proportion	0.4455 – 0.5743	202
S1 strict class-level	0.4857 proportion	0.4571 – 0.7143	35
Unassigned share, strict corpus	0.7755 proportion	—	147
Effective number of compounds (1/HHI), loose primary	22.5 compounds	—	159
Scenarios with S1 < 0.50	9 scenarios	—	10

Next: Manually review the 50-paper precision sample and the 114 unassigned strict papers, extend the dictionary with what they reveal, and cross-check recall against PubMed MeSH (Drug Repositioning + Leukemia, Myeloid, Acute) before re-estimating S1.

Caveat · open

Only 33 of the 147 strict-corpus papers (22%) were matched to a dictionary drug, and the rank-5 slot is a tie, so the primary estimate is fragile. Drug assignment relies on title/abstract keyword matching, and no manual precision check has been done yet. Results also depend on the frozen dictionary and on the boundary drawn for "non-oncology".

Claims

scripts/aml-repurposing-1/analysis.py
Europe PMC counts 2015-2025: all AML records; strict AML+repurposing query hits; strict papers kept after filters; loose (drug-in-title) papers kept.
Passed· 5.8s · 15 h ago
scripts/aml-repurposing-1/analysis.py
Primary top-5 share of drug-assigned papers (S1): strict 0.4242, loose 0.4088; top-1 share and effective number of compounds.
Passed· 6.5s · 15 h ago
scripts/aml-repurposing-1/analysis.py
Sensitivity S1: with ATRA, class-level, preclinical-only scenarios.
Passed· 13.8s · 15 h ago

Literature