Identifying a minimal set of single nucleotide polymorphisms to classify the geographic origin of a P. falciparum sample from the pf3k database

Genetic sequencing of malaria parasites has the potential to become an important tool in routine surveillance efforts for the control and eradication of malaria. For example, characterizing the epidemiological connectivity between different populations by assessing the genetic similarity of their parasites can offer insights for national malaria control programs and their strategic allocation of interventions. Despite the increase of whole-genome sequencing of malaria parasites, the development of a small set of single nucleotide polymorphisms (SNPs), often referred to as a barcode, or a panel of amplicons remains programmatically relevant for large-scale, local generation of genetic data. Here, we present an application of a machine-learning method to classify the geographic origin of a sample and identify a small set of region-specific SNPs. We demonstrate that this method can automatically identify sets of SNPs which complement the currently targeted loci from the malaria scientific community. More specifically, we find that many of these machine-learned SNPs are near known and well-studied loci such as regions and markers linked to drug resistance, while also identifying new areas of the genome where function is less characterized. The application of this technique can complement current approaches for selecting SNP locations and effectively scales with an increase in sample size.