01 / The question
Did African malaria remedies leave a chemical trace?
Two frontline antimalarials came out of plants that people already used against fever. We asked whether the historical record is predictive in general, and tested it against — species in — botanical families.
Every number on this page is read from the project's own output files. None is typed in by hand.
02 / The premise
Two of the drugs that work came out of plants
The idea that traditional remedies encode real pharmacology is not naive. It has produced a frontline drug twice. What nobody counts is the failures.
Quinine
Isolated from Cinchona bark, long used against fever in Peru.
Artemisinin
Isolated from Artemisia annua, found by reading a fourth-century Chinese fever text.
of malaria deaths
occur in the WHO African Region, which is where the plant knowledge we drew on was recorded.
WHO World Malaria Report
03 / The hypothesis
Written so that it could fail
Plants that African sources record as malaria remedies contain a higher share of compounds predicted active against Plasmodium falciparum than other medicinal plants from the same book.
The test was fixed first
The statistic, the permutation scheme and all three sensitivity analyses were written down before any result was computed.
The circularity was removed
Malaria plants were often assayed because of their traditional use. Every compound the model trained on is excluded from scoring.
Labels shuffle within families
Related plants share chemistry, so the null model permutes inside each botanical family rather than across all of them.
04 / Method
From papyrus to prediction
Every recipe keeps its page and plate number, so any claim about the papyrus can be checked against Joachim's 1890 translation.
The Ebers Papyrus never names malaria. Of — recipes, exactly one is explicitly for fever-heat, which is why the statistics rest on Watt & Breyer-Brandwijk (1932) instead.
05 / The papyrus
What the papyrus actually treats
Malaria was never the papyrus's subject. Read whole, its — remedies are mostly for the belly, the skin and the eyes. Fever gets —.
Each remedy is counted once, under the body system its stated purpose names. Hover a bar for the first remedy of that kind in papyrus order.
Honey, copper and lead
Honey appears in — remedies, copper salts (verdigris) in — and red lead in —. Honey's antibacterial action on wounds and copper's on microbes are both well documented today. None of it is a plant, so none of it reaches a pipeline built on plant chemistry, and the wound remedies, not fever, are where a second test belongs.
The worm cure the database forgot
Plate — prescribes pomegranate bark, pounded in beer and drunk, against worms that "die from no other remedy". That bark carries the pelletierine alkaloids, which European doctors gave against tapeworm into the twentieth century. The open database lists — compounds for pomegranate, and pelletierine —.
06 / Instrument check
The model knows a real antimalarial
Before trusting a null result, check that the instrument reads true. Plant natural products sit low. The two reference drugs sit far right.
The model is trained on whole-cell screening data: thousands of plates like this one, where a handful of wells come back active and most do not.
Trained on — ChEMBL assays against P. falciparum, scored by five-fold scaffold-grouped cross-validation. Random splits leak analogue series between train and test and would have inflated this. Quinine and artemisinin appear only as controls and are excluded from every analysis.
07 / Result
No signal, in any version of the test
Three analyses, each an estimate with an interval. All three cross zero, and the effect collapses once thinly documented species are removed.
Positive means malaria-use plants score higher. The shaded band is the smallest difference this study could reliably detect, —, about 0.52 standard deviations. We can rule out an effect larger than that band. We cannot rule out a smaller one.
The drift was noise
Requiring at least five compounds per species drops the difference to —. The small positive in the primary analysis came from species with one to four recorded compounds, where a mean is almost meaningless.
Research effort is not the explanation
Better-studied plants do score slightly higher (Spearman —), but malaria plants are not better studied: a median of — compounds against — for the rest.
08 / Species
The two groups overlap almost entirely
The same result, one point per species. Hover any point. Filter by family, or drop the species whose score rests on too little chemistry and watch the gap close.
09 / The finding that matters
Most of these plants have no chemistry
This is not a gap in the pipeline. It is a gap in the world's published chemistry, and it reframes the null result above.
Three jars on a rack built for twelve. That is roughly the ratio: most of these plants have never had their chemistry described at all.
Searched against LOTUS via Wikidata. Median for species that have any record at all: 13 compounds.
10 / A second tradition
Rich chemistry, same answer
If missing data hid the signal, a source with plenty of chemistry should show it. We repeated the test on al-Heravi's tenth-century Persian materia medica, Al-Abniah, which names malaria-like fevers outright. The design was committed before a single score was computed.
Same model, score and within-family permutation as Fig.03. The top row is the African result, shown for comparison. The shaded band is the smallest difference this test could reliably detect, —, about — standard deviations against — for the African test.
Fevers that come and go
Unlike Ebers, Al-Abniah names Febris quartana, the tertian fever ghibb, and rubs applied "before the fever attack", which is the rhythm of malaria. The — scored species recommended for those fevers sit no higher than the rest.
What this does to the data-gap story
Missing chemistry is still real, and it still limits the African test. But here the chemistry is there and the answer does not change. The likelier reading for both traditions is the one in the limits below: a remedy could ease a fever without touching the parasite.
11 / One lead
The plant still worth a bench test
Of — untested compounds from malaria-use plants, — score above 0.5. One plant stands out for a historical reason rather than a statistical one.
Cassia abbreviata
The only plant in Watt (1932) carrying a clinical observation rather than a report of custom. A physician in East Africa found a root decoction helped in blackwater fever, and a fluid extract was manufactured and prescribed.
Highest-scoring malaria plant with more than ten compounds: mean —, with — of — predicted active.
Why we are not calling it a hit
Its top-scoring compounds share an identical predicted score, which means they are near-identical structures. This is a handful of scaffolds, not eleven independent findings.
It needs an assay, not a press release.
Read by hand from 63 malaria and blackwater-fever statements. Nine further statements were excluded with a recorded reason, including the belief that the fever tree causes malaria. Paraphrased, because the 1932 text is probably still in copyright.
12 / Limits
What it would take to answer this properly
More chemistry may not change the answer
— of these Southern African species have no compound on record, and that gap is real. But the Persian test had a median of — scored compounds per species and still found nothing, so filling the gap is no guarantee of a signal.
West Africa is missing
Dalziel's 1937 survey has no open full text, so the record covers Southern and Eastern Africa alone. Twenty species cannot carry this question.
A remedy need not kill the parasite
Many recorded uses induce sweating or vomiting. A plant that eased a fever could earn its reputation without ever touching Plasmodium. Two traditions, a thousand years and a continent apart, now point the same way.
Ebers Papyrus via Joachim (1890, public domain)
Watt & Breyer-Brandwijk, The Medicinal and Poisonous Plants of Southern Africa (1932)
al-Heravi, Al-Abniah, via A. Achundow (1893, public domain)
GBIF Backbone Taxonomy · LOTUS via Wikidata · ChEMBL target CHEMBL364