Depth and evenness of sequencing coverage are considered potential indicators of genome assembly quality. In plastid genomics, where new data generation has outpaced the development of assembly quality indicators, these coverage metrics could offer insights into the quality of plastomes of different sizes, structures, or taxonomic origins. However, the variation of sequencing depth and evenness among archived plastid genomes, their variability between genome partitions, and any association with methodological factors have yet to be evaluated. This study explores the variation of sequencing depth and evenness across a sample of publicly accessible plastid genomes in relation to their genome structure, assembly quality, and methodological provenance using uni- and multivariate statistical analyses. We also evaluate whether sequencing evenness in plastid genomes is biased by phylogenetic signal and assembly software choice, and whether more uniformly distributed input sequence data improves plastome assembly quality. Our results indicate significant differences in sequencing depth across the four structural partitions and between the coding and non-coding regions of plastid genomes, a significant correlation between sequencing evenness and the number of ambiguous nucleotides, and a significant difference in sequencing evenness between sequencing platforms. However, we also find that different covariates representing additional, lesser explored factors often show a similar, if not greater, explanatory power for the coverage variation. No indications of phylogenetic or software choice bias on sequencing evenness and only weak indications of phylogenetic bias among the assembly quality metrics are detected, suggesting that our study results represent genuine patterns. We also find that normalizing the distribution of the input sequence data before plastome assembly may improve assembly accuracy. Taken together, these findings highlight that many public plastid genomes derive from sequence data with highly variable depth and evenness, and that this variation is influenced, at least partially, by genome structure as well as methodological factors.
