Fig. 3

Differences among four QPX strains and with the reference genome. The principal component analysis is based on 100 annotated functional sequences that are selected after MG-RAST quality-control checks of the annotation process and sequence similarity search on 5 QPX libraries that include four QPX strains and the reference genome. Feature selection is done with an identity score (maximum of 1 mismatch) for annotating the QPX contigs against protein domains, an alignment length score (similarity ≥ 80%) that represent the coverage length and similarities between estimated alignments, an e-value score (10^-10) for predicted functional similarities, and the number of times an annotated contigs is identified. The arrows are eigenvectors, they represent linear trends in the data