tools/coverage_map.md

# Literature coverage map (2026-08-05)

How much of the FINDABLE literature does the challenge pool actually hold? Nobody had
established the denominator, so 'find more papers' was untargeted. This measures it.

## Method

For each hypothesis-defining PubMed query, retrieve the full ID list and intersect with the
291 PMIDs the pool has captured. Coverage = pool ∩ slice / slice.

| slice | PubMed hits | pool has | coverage | unexamined |
|---|---|---|---|---|
| microglia + Abeta phagocytosis | 1199 | 43 | 4% | 1156 |
| ABCA1 + APOE (any) | 877 | 63 | 7% | 814 |
| microglia + lipid droplet | 334 | 66 | 20% | 268 |
| ABCA1 + astrocyte | 140 | 30 | 21% | 110 |
| ABCA1 + Alzheimer | 330 | 84 | 25% | 246 |
| APOE4 + microglia + phagocytosis | 72 | 32 | 44% | 40 |
| APOE4 + lipid droplet | 56 | 34 | 61% | 22 |

Queries are recorded verbatim in the JSON alongside this file so the numbers are reproducible.

## Reading it

Tight, clause-specific queries are well covered (APOE4+lipid droplet 61%, APOE4+microglia+
phagocytosis 44%). Broad slices are barely touched (microglia+Abeta phagocytosis 3.6% of 1199).
That is the expected shape: agents searched the hypothesis wording, not the surrounding
mechanism literature. The gap is real but it is mostly OFF-CLAUSE, so raw unexamined counts
overstate the opportunity.

## Triage that survives that caveat

179 unexamined papers sit at the intersection of >=2 hypothesis-defining slices; 140 are
primary research; 65 are 2018 or later. Scored on endpoint vocabulary:

| score | slices | PMID | year | title |
|---|---|---|---|---|
| 11 | 2 | 37546938 | 2023 | APOE4/4 is linked to damaging lipid droplets in Alzheimer's microglia. |
| 10 | 2 | 39456234 | 2024 | Inhibiting the Cholesterol Storage Enzyme ACAT1/SOAT1 in Myelin Debris-Treated Microglia |
| 9 | 2 | 38525390 | 2024 | Investigating the Effect of Aspirin on apoAI-Induced ATP Binding Cassette Transporter 1  |
| 7 | 3 | 34744007 | 2022 | Amyloid beta increases ABCA1 and HMGCR protein expression, and cholesterol synthesis and |
| 7 | 2 | 39840463 | 2025 | Intermittent Fasting Ameliorates β-Amyloid Deposition and Cognitive Impairment Accompani |
| 7 | 2 | 37564436 | 2023 | 24-Hydroxycholesterol Moderates the Effects of Amyloid-β on Expression of HMG-CoA Reduct |
| 7 | 2 | 29605487 | 2018 | Association between ABCA1 gene polymorphisms and the therapeutic response to donepezil t |
| 6 | 3 | 38699303 | 2024 | Comparative gene regulatory networks modulating APOE expression in microglia and astrocy |
| 6 | 3 | 37488092 | 2023 | Inhibition of Sirt2 Decreases ApoE Secretion in Astrocytes and Microglial Cells. |
| 6 | 3 | 30474860 | 2018 | HDL mimetic peptides affect apolipoprotein E metabolism: equal supplement or functional  |
| 6 | 2 | 42554945 | 2026 | GPR146 Deficiency Enhances Microglial Phagocytosis and Blood-Brain Barrier-Associated Ma |
| 6 | 2 | 40661651 | 2025 | Inhibition of Acyl-CoA: Cholesterol Acyltransferase 1 Promotes Shedding of Soluble TREM2 |
| 6 | 2 | 39612244 | 2024 | Microglial APOE3 Christchurch protects neurons from Tau pathology in a human iPSC-based  |
| 6 | 2 | 38191520 | 2024 | Sex-dependent effect of sublethal copper concentrations on de novo cholesterol synthesis |
| 6 | 2 | 29323721 | 2018 | Brain-derived neurotrophic factor modulates cholesterol homeostasis and Apolipoprotein E |
| 4 | 4 | 39345217 | 2024 | Deletion of miR-33, a regulator of the ABCA1-APOE pathway, ameliorates neuropathological |
| 4 | 3 | 34949639 | 2021 | Neuronal ROS-induced glial lipid droplet formation is altered by loss of Alzheimer's dis |
| 4 | 3 | 29563219 | 2018 | Small molecule inducers of ABCA1 and apoE that act through indirect activation of the LX |
| 4 | 2 | 42325211 | 2026 | Exploration of precision coregulator TR-FRET identifies diverse signatures for LXR ligan |
| 4 | 2 | 41000830 | 2025 | In search of nonlipogenic ABCA1 inducers (NLAI): precision coregulator TR-FRET identifie |
| 4 | 2 | 31441725 | 2019 | Integrated Genomic Analysis Revealed Associated Genes for Alzheimer's Disease in APOE4 N |
| 4 | 2 | 31168344 | 2019 | The effects of aerobic training before and after the induction of Alzheimer's disease on |
| 3 | 3 | 34045954 | 2021 | Dichlorodiphenyltrichloroethane Impairs Amyloid Beta Clearance by Decreasing Liver X Rec |
| 3 | 2 | 42079117 | 2026 | Epigenetically constrained astrocyte states underlie prefrontal cortex vulnerability in  |
| 3 | 2 | 40334811 | 2025 | Astrocyte lactoferrin deficiency affects the construction and function of spinal neurons |
| 3 | 2 | 39345559 | 2024 | The mechanism and consequences of amyloid-β modulating thiamine pyrophosphokinase-1 expr |
| 3 | 2 | 38856177 | 2024 | Lysosomal cholesterol accumulation in aged astrocytes impairs cholesterol delivery to ne |
| 3 | 2 | 36138870 | 2022 | Application of Micro-Western Array for Identifying Different Serum Protein Expression Pr |
| 3 | 2 | 33897406 | 2021 | Interplay Between Microglia and Alzheimer's Disease-Focus on the Most Relevant Risks: AP |
| 3 | 2 | 30010117 | 2018 | ABCA7 Downregulation Modifies Cellular Cholesterol Homeostasis and Decreases Amyloid-β P |

Extracted by k-dense from this queue: 39345217 (M1H2 + M3H1), 34949639 (M3H2).
37546938 was SKIPPED - it is the bioRxiv preprint of 38480892, already pooled.

## Capture-recapture, and why I do not trust it

Treating each agent's search as a capture occasion: 292 distinct PMIDs, 167 singletons,
73 doubletons, Chao1 = 479 (95% CI 402-557), implying ~61% completeness. BUT leave-one-agent-out
validation fails badly - holding out nakos predicts 310 missing papers and the agent actually
contributed 43 unique. Median pairwise overlap is 0.51x of the independence expectation, so
agents searched DIFFERENT niches and Chao1's homogeneity assumption is violated. The 61% figure
is reported here only so nobody re-derives it and trusts it. Use the per-slice coverage table
instead: it is measured, not modelled.