PX1 polymorphisms and ex-vivo antimalarial susceptibility in Uganda

Among source-reported Ugandan Plasmodium falciparum isolates carrying high-major-allele PIN or LMD PX1 calls, do ex-vivo antimalarial susceptibility measurements differ, overall and within K13 wild type?

The original study investigated selection and drug susceptibility patterns for the PX1 haplotype in Ugandan Plasmodium falciparum. The paper reports associations for DHA, lumefantrine and mefloquine, while its ex-vivo RSA comparison was not statistically significant. Cathedral applies a separate unadjusted analysis to the accessible phenotype workbook and makes the differences from the original cohort and model explicit.

Primary publication: Niaré K et al., Nature Medicine (2026), “Emergence and spread of Plasmodium falciparum PX1 polymorphisms associated with decreased susceptibility to antimalarials in Uganda.” Read the publisher record. The WHO marker compendium is provided for context; it does not classify PX1/PIN as a validated marker.

In the accessible source workbook and under Cathedral's available filters, PIN-labelled records had higher median IC50 values than LMD-labelled records for dihydroartemisinin, lumefantrine and mefloquine. The RSA comparison did not show a statistically clear difference. These are associations in published ex-vivo measurements, not evidence of causation, patient treatment failure, or a validated resistance marker.

The directions are consistent with the paper's reported results, but the sample counts differ. Cathedral lacks numeric per-sample read depth and the sample-linked covariates needed to reproduce the paper's 465-sample cohort and adjusted mixed-effects model.

Assay-specific complete cases from the accessible workbook. The difference is PIN median minus LMD median. These comparisons are unadjusted.

p-values use the implemented tie-corrected normal approximation with continuity correction. Bootstrap intervals use 4,000 within-group resamples; Benjamini-Hochberg q-values are adjusted across the four assays. Exact values, U statistics, and missingness are in the CSV and JSON artifacts.

This secondary analysis uses the same filters and a separate four-assay adjustment. It is not the exact paper cohort.

1. Inputs are the publisher's `Source_dataset1` workbook and WHO's marker compendium. The manifest lists source URLs, checksums, retrieval date, and reuse notes. Raw workbooks are not distributed here.
2. Filter to source-labelled PIN or LMD, WSAF greater than 0.9, and K13 categories used by the authors' analysis code. No numeric per-sample sequencing depth was available, so the paper's at-least-25x criterion could not be applied. Assays use their own complete cases.
3. Compare PIN and LMD with a two-sided Mann-Whitney U / Wilcoxon rank-sum test using a tie-corrected normal approximation with continuity correction. The effect summary is the PIN-minus-LMD median difference.
4. Bootstrap percentile intervals use 4,000 resamples within each group and deterministic seed 20261029. Benjamini-Hochberg adjustment is applied within each four-assay family. K13 wild type is a separate exploratory family.
5. Run and test the pipeline with the reproduction instructions. No Supabase, Modal, or API key is required.

This is not the original paper's mixed-effects model.

The article reports 465 depth-qualified samples with at least 25x sequencing depth and at least 90% major-allele support. In K13-wild-type samples, published medians were DHA 3.7 nM (PIN n=110) versus 1.8 nM (LMD n=240); lumefantrine 14.3 nM (PIN n=110) versus 6.2 nM (LMD n=240); and mefloquine 17.1 nM (PIN n=109) versus 11.0 nM (LMD n=236). The paper reports no significant ex-vivo RSA difference (n=169 overall).

Its adjusted model controls for site, parasitemia, COI, year and K13 genotype. Published narrative values include DHA marginal means 3.1 nM (PIN) and 2.3 nM (LMD), p=3.0e-8, Cohen's d=0.44, and lumefantrine means 14.2 nM and 8.8 nM, p=0.0025. The PIN term explained 12.3% of phenotypic variance for DHA in K13 wild type, 4.9% for lumefantrine, and 6.5% for mefloquine. The adjusted RSA association was not significant (p=0.13). Interaction tests were p>0.65 for lumefantrine, mefloquine and RSA; the DHA interaction was p=0.061.

Cathedral's current K13-wild-type DHA comparison is PIN n=115, median 3.70 nM, versus LMD n=236, median 1.77 nM. Similar direction does not resolve the cohort mismatch. The unadjusted median difference and adjusted estimated marginal means are different estimands; they should not be compared as equivalent effect estimates. The full extended-data source workbook is linked by the paper, but exact confidence-interval bounds and all table cells have not been independently extracted here.

Read the detailed comparison and the exact reproduction gaps.

Partial reproduction; further validation required

1. The paper's mixed-effects model is not reproduced: the retrieved source workbook lacks COI (complexity-of-infection) values, and row-level numeric read-depth data are not present. The original model included site, parasitemia, COI, year and K13.
2. The author's R script also consumes separate phenotype, parasitemia and COI files; this analysis uses only the article's published Source_dataset1 table and cannot recreate undocumented joins or full model provenance.
3. The article's data are observational ex-vivo susceptibility measurements, not randomized clinical treatment outcomes.
4. Applying the accessible source workbook with available filters does not reproduce the article's exact analytic sample: K13-wild-type DHA counts are 115 PIN and 236 LMD here, versus 110 and 240 reported. Numeric read depth is absent, so the paper's >=25x inclusion rule cannot be verified. Treat this as an independent partial reanalysis, not exact sample-level reproduction.
5. This is a partial reproduction of already published comparisons, not a novel discovery or independent biological replication.
6. Source workbook reuse and redistribution are constrained by the article's stated CC BY-NC-ND license; obtain dataset-specific permission before using these artifacts commercially or redistributing the workbook or adapted content.
7. The exact original mixed-effects model and PX1-K13 interaction have not been reproduced.
8. Independent review by a malaria researcher is still needed.

The next step is to request the depth-qualified cohort mapping, sample-linked covariates, and model specification from the authors, then have a malaria researcher review the reproduction. The author data-request draft is here; it has not been sent.

The WHO compendium lists validated K13 markers including C469Y and A675V. Cathedral's analysis does not establish PX1/PIN as a WHO-validated marker.

Aggregate outputs, provenance, executable analysis code, tests, and a Colab-ready notebook are served as static files from this repository. No database session is required to read this case study. The publisher's and WHO's source workbooks are not mirrored; follow official links and respect their terms.

• Final report
• Machine-readable summary
• Phenotype comparisons (CSV)
• Aggregate results figure (SVG)
• Source manifest and checksums
• Analysis notebook
• Analysis script
• Source fetch script
• Reproduction instructions
• Analysis tests
• Comparison with the original paper
• Draft author data request (not sent)

Analysis generated 2026-10-29 · seed 20261029 · source files excluded from publication