Additional file 4 of Development of a new barcode-based, multiplex-PCR, next-generation-sequencing assay and data processing and analytical pipeline for multiplicity of infection detection of Plasmodium falciparum

Additional file 4: Figure S4. StrainRecon reconstruction on laboratory-mixed samples. Ground truth of barcodes (first row in each figure), the MAP estimate of the reconstruction matrix M and mixture vector v (second row), and the mean (third row) and standard deviation (fourth row) of the posterior density of candidate (M, v) solutions. Each block contains one row for each strain, ordered by decreasing frequency from top, with a SNP cell color ranging from purple (fraction of 0) to bright yellow (fraction of 1). Each reconstruction is shown as the input parameter n (of the number of strains) is varied from n = 1 and 4 (row-major order) to be reconstructed in StrainRecon on the B_24NOPA_DBS sample. The dominant strain is captured perfectly and with high confidence by considering the posterior statistics. The algorithm has difficulty reconstructing the other two less-prevalent strains, since their target range of < 4% is low relative to the experimental noise levels seen in the pipeline.