An AI model read the abstract of every ALS paper in PubMed and recorded each claim: what was studied, what it did, in what kind of experiment, with the sentence it came from. Put side by side, the claims show what human trials found, what works in animals but has not been tried in people, and where studies disagree.
Read these as leads, with the evidence attached. Every claim links to its paper and shows the exact sentence it rests on, and that sentence was checked against the abstract. The label on each claim (reduces, no effect, and so on) was assigned by the model and has not been graded by experts. Only abstracts were read. Most results in cells and animals do not carry over to people.
Treatments with claims from human trials in three or more papers. For each outcome, the green bar counts papers reporting an improvement and the red bar papers reporting no effect. Several papers can describe the same trial, so this counts reports, not trials.
| Treatment | Papers | By outcome: improved / no effect |
|---|
Drugs that papers report as reducing disease measures in cells or animals (at least four papers, two in animals), with no human-trial claim in the literature and no matching trial in the registries this atlas collects. Laboratory tool compounds and whole drug classes are filtered out.
| Drug | Lab papers | In animals | Against | Latest | Reported to reduce | Example |
|---|
For ten things widely implicated in ALS, the drugs and cell therapies reported in two or more lab papers to reduce them. “Not tried” means no human-trial claim and no registry trial we hold.
Subject and outcome pairs where at least three human papers report an effect or link and at least three report none. Disagreement is often real: different patients, doses, or stages of disease.
| Subject | Outcome | Papers: effect or link | Papers: none |
|---|
Every gene, drug, biomarker and exposure with claims in three or more papers. Choose one to see its evidence by kind of study, then read the claims and their sentences.
Choose something from the list.