To increase the stringency of SNP identification, the database was queried for SNPs identified by samtools, and only SNPs identified by both methods are included in the final analysis. Two complete genome sequences of A. marginale strains from the United States (Florida and St. Maries, Idaho) and one this website of A. marginale subspecies centrale (Israel) are available [14], [26] and [27]. We analyzed high-throughput sequencing data from the Roche/454 instrument on 10 U.S. A. marginale strains, including the previously genome-sequenced Florida and St. Maries strains as controls. Including Florida and St. Maries strains enables a comparison to be made between the new pyrosequencing

data and data obtained using Sanger sequencing. We included in this comparison a second Florida strain (Okeechobee) and

a second Idaho strain (South Idaho). We also included a Florida relapse strain derived from a persistently infected animal after 129 days of infection, to examine genome changes over a short time period. The initial analyses compared the original genome sequences with the new pyrosequencing data. This was done by aligning individual pyrosequenced reads with the completed genomes using Mosaik, with visualization of the finished PI3K inhibitor alignments using Artemis. To deal with the known problem of multiple repeats in these genomes, the alignment parameters were set to allow reads to align at multiple different positions in the genome, if this was necessary. A typical result showing alignments with msp2 and msp3 genes is shown in Fig. 1. The top panel shows alignment of Florida strain pyrosequencing data with a region of the Florida genome containing an msp2/msp3 gene pair (AMF_871/872). The reads align over the complete msp2 and msp3 regions, as expected. In the middle panel, a comparison is made much between the same Florida strain pyrosequencing

data but with a region of the St. Maries, Idaho strain genome encompassing the msp2/msp3 gene pair AM1344/1345. In this case, the previously obtained genome data shows that AM1344 has an exact match (100% identity) with an msp2 copy in the Florida strain genome, but the closest match of the St. Maries msp3 copy AM1345 is to an msp3 copy in the Florida strain with only 78% identity ( Table 1). This is revealed by a gap in the aligning sequence reads over the central (hypervariable) region of AM1345, but no gap over AM1344. The lowest panel shows an extreme case where neither the msp2 (AMF_1018) nor the msp3 (AMF_1019) pseudogene from the Florida strain aligns with reads from St. Maries. Comparison of the two genome sequences reveals closest matches between the two genomes of 91% for AMF_1018 and 55% for AMF_1019. This analysis was conducted for all msp2 and msp3 copies in the three genomes, A. marginale (Florida strain), A. marginale (St.