genomic variation - Pf

• The reference standard is the MalariaGEN P. falciparum open variation resource (currently Pf8; 41267740),
  Pf8: An open dataset of Plasmodium falciparum
  
• queried through Pf-HaploAtlas (https://pf-haploatlas.streamlit.app) for a first-pass per-protein answer, and re-analysed yourself with the malariagen_data API if you need statistics.

8-26 Analysis approach

1. The database layer

Primary variation catalogue — MalariaGEN Pf8. Pf8 curates genome variation from over 33,000 samples from 99 partner studies and 122 locations spanning more than 50 years, with open access to raw sequencing data and genotypes at over 12 million genomic positions, and for the first time includes copy-number variation calls in drug-resistance-associated genes alongside deletion calls in the genes underlying rapid diagnostic test failures (PMID 41267740, DOI). It also ships derived datasets, interactive web applications for exploring variation in over 5,000 genes, an updated Python package for data access and analysis, and open analysis notebooks. The predecessor Pf7 (~20,000 samples, 6 M SNPs/indels) is still widely cited and is the substrate of most published gene-level diversity papers (PMID 36864926, DOI).

Fastest route to your specific question — Pf-HaploAtlas. It provides analysis-ready data catalogues and visualisations of amino-acid haplotypes for all 5,102 core P. falciparum genes, covering 16,203 samples from 33 countries between 1984 and 2018, and allows download of the underlying amino-acid haplotype data for further analysis (PMID 39565917, DOI). For a protein of interest this gives you, in minutes: number of distinct amino-acid haplotypes, per-residue mutation frequencies, and their spatiotemporal trajectories.

Annotation/reference layer — PlasmoDB (VEuPathDB). Use it for the Pf3D7 reference coordinates, ortholog groups and functional evidence you will join your variant calls to (PMID 37953350, DOI).

2. Best-practice rules before you build the pipeline

• Check your gene is in the core genome first. Subtelomeric and hypervariable families (var, rifin, stevor) are not reliably genotyped from short reads; if your protein belongs there, sequence-based conservation must come from long-read assemblies instead (PMID 39920908, DOI).
• Do not run SNP-only analyses. Indels are exceptionally abundant in P. falciparum, more common than SNPs, and are the dominant mode of polymorphism within the core genome (PMID 27531718, DOI). Low-complexity, AT-rich regions are also where genotype error concentrates — apply the MalariaGEN callable/core mask rather than your own.
• Include copy number, not just point mutations. In systematic resistance selection, 159 gene amplifications and 148 nonsynonymous changes in 83 genes were associated with resistance acquisition, with gene amplifications accounting for one-third of resistance-acquisition events (PMID 29326268, DOI). A target that is invariant at the amino-acid level can still escape by amplification.
• Filter for monoclonality. Polyclonal infections generate heterozygous calls; standard practice is to restrict haplotype-level statistics to monoclonal samples (F_WS threshold), as done for Pfs230 where 1,312 of 1,471 sequences passed quality filtering and 496 haplotypes were counted among 718 monoclonal sequences (PMID 42226229, DOI).
• Mind the sampling bias. Coverage is dominated by Africa and SE Asia and is temporally uneven; a "rare" variant may simply be unsampled. Stratify by country and year rather than pooling.

3. Recommended metric panel for "is my protein variable?"

The Pfs230 study is a clean template for exactly this workflow on a candidate target (PMID 42226229). The dhfr/dhps analysis shows the temporal-trajectory version, which is what you want for resistance forecasting (PMID 40247054, DOI).

4. Two orthogonal layers you should add for drug development

Natural variation alone is not sufficient for target de-risking.

1. Essentiality/fitness cost — saturation transposon mutagenesis generated >38,000 mutants and defined mutability and fitness costs for over 87% of genes, identifying 2,680 of 5,399 genes as essential for optimal asexual blood-stage growth in vitro (PMID 29724925, DOI). A target under strong purifying selection and essential is the best case.
2. Experimental resistome — in vitro evolution with whole-genome analysis (IVIEWGA) tells you which escape mutations the parasite can actually reach under drug pressure, which natural variation cannot predict (PMID 29326268).

Practical decision rule: low π, negative Tajima's D, dN/dS well below 1, no high-frequency non-synonymous residues in the binding pocket, essential by piggyBac, and no CNV signal → good, conserved target. High haplotype diversity with positive Tajima's D or geographically structured non-synonymous variants in the pocket → treat as resistance-prone.

Literature

PMID 41267740 — MalariaGEN. Pf8: an open dataset of genome variation in 33,325 worldwide samples. Wellcome Open Res 2025. DOI

This release curates whole-genome variation from 33,325 samples across 99 studies and 122 locations, providing open genotypes at more than 12 million genomic positions. It is the first MalariaGEN release to add copy-number variation calls in additional drug-resistance genes alongside the existing CNV and diagnostic-deletion calls. It ships derived datasets, interactive web apps covering more than 5,000 genes, an updated Python access package and open analysis notebooks, making it the largest open eukaryotic genome-variation dataset available.

PMID 39565917 — Lee C et al. Pf-HaploAtlas: an interactive web app for spatiotemporal analysis of Plasmodium falciparum genes. Bioinformatics 2024. DOI

The authors built a web application that converts MalariaGEN whole-genome data into analysis-ready amino-acid haplotype catalogues for all 5,102 core P. falciparum genes. It covers 16,203 samples from 33 countries collected between 1984 and 2018 and visualises haplotype frequencies across space and time without requiring bioinformatics expertise. The underlying haplotype tables are downloadable, so it doubles as a data source for downstream custom analyses.

PMID 36864926 — MalariaGEN. Pf7: an open dataset of genome variation in 20,000 worldwide samples. Wellcome Open Res 2023. DOI

Pf7 released genotypes at 6 million SNPs and short indels from over 20,000 samples, including dried blood spots sequenced after selective whole-genome amplification, which required new CNV genotyping methods. It identified newly emerging resistance mutations in Southeast Asia and documented heterogeneity in resistance patterns within Africa and the Indian subcontinent. It also characterised C-terminal csp variation relative to the sequences used in the RTS,S and R21 vaccines.

PMID 27531718 — Miles A et al. Indels, structural variation, and recombination drive genomic diversity in Plasmodium falciparum. Genome Res 2016. DOI

Using deep sequencing of parents and progeny from genetic crosses, Mendelian error rates were used to benchmark genotype accuracy for SNPs, indels and complex polymorphisms genome-wide. The key finding is that indels outnumber SNPs and are the dominant polymorphism class in the core genome, which reshapes how variant-calling pipelines must be designed. The dense marker set also yielded the first estimates of non-crossover rates and conversion tract lengths, and revealed meiotic recombination inside drug-resistance-associated CNVs.

PMID 29724925 — Zhang M et al. Uncovering the essential genes of Plasmodium falciparum by saturation mutagenesis. Science 2018. DOI

Transposon mutagenesis exploiting the AT-rich genome produced more than 38,000 mutants and quantified mutability and fitness cost for over 87% of genes. Of 5,399 genes, 2,680 were defined as essential for optimal asexual blood-stage growth in vitro, including roughly 1,000 conserved genes of unknown function. The dataset provides the mutagenesis index/fitness scores now routinely used to prioritise drug targets.

PMID 29326268 — Cowell AN et al. Mapping the malaria parasite druggable genome using in vitro evolution and chemogenomics. Science 2018. DOI

Genome analysis of 262 parasite lines resistant to 37 chemically diverse compounds mapped the experimental resistome. Resistance was attributable to 159 gene amplifications and 148 nonsynonymous changes across 83 genes, with amplification accounting for about a third of all resistance events. New target–inhibitor pairs were identified, including thymidylate synthase, farnesyltransferase and a dipeptidyl peptidase.

PMID 40247054 — White NFD et al. Disparate co-evolution and prevalence of SP resistance alleles at dhfr and dhps across Africa. Sci Rep 2025. DOI

Using Pf7, the authors mapped dhfr/dhps haplotype frequencies across 22 African countries and tracked change over time in four of them. The dhfr N51I/C59R/S108N triple mutant remained dominant continent-wide, while dhps showed much greater haplotype diversity and spatial heterogeneity, with K540E nearly absent in Madagascar despite low genome-wide differentiation from mainland East Africa. Twelve further highly differentiated SNPs were identified, including aat1 and a candidate novel resistance locus ~20 kb downstream of mdr1.

PMID 42226229 — Kisambale AJ et al. Genetic diversity of Pfs230 in four East African countries. Malar J 2026. DOI

This is a worked example of the target-conservation workflow applied to a transmission-blocking vaccine candidate using MalariaGEN Pf7 data from Ethiopia, Kenya, Tanzania and Uganda. From 1,471 retrieved sequences, 1,312 passed filtering; nucleotide diversity was low, dN/dS was 0.43 and Tajima's D was negative in all countries, significantly so in Kenya and Tanzania. PCA and F_ST showed no population differentiation, supporting Pfs230 as a conserved, low-escape-risk candidate.

PMID 39920908 — Lawton JG et al. Alignment-free comparative genomics reveals strain-transcendent P. falciparum antigens. Infect Genet Evol 2025. DOI

k-mer comparison of RIFIN and STEVOR proteins across 18 long-read P. falciparum assemblies quantified family-wide diversity, with mean amino-acid k-mer similarity of 49.5% and 61.1% respectively, dropping to 25.1% and 20% in hypervariable regions. Despite this, 11 RIFINs and five STEVORs were conserved across strains above expectation and syntenic with genes in related Plasmodium species. AlphaFold predictions showed that RIFIN receptor-binding regions are structurally more conserved than their sequences imply — a caution that sequence variability alone can overstate epitope/pocket variability.

PMID 37953350 — Alvarez-Jarreta J et al. VEuPathDB in 2023. Nucleic Acids Res 2024. DOI

VEuPathDB integrates genomes and pre-analysed omics data for more than 600 eukaryotic pathogens, vectors and hosts, including PlasmoDB. It couples over 3,000 pre-analysed datasets with orthology prediction via OrthoMCL, a search/visualisation layer and a Galaxy workspace for private data. This release added new data types, a redesigned MapVEu spatial interface and updated infrastructure.