Fig. 1From: Relative model selection of evolutionary substitution models can be sensitive to multiple sequence alignment uncertaintyPercent of stable and unstable (defined in Table 1) datasets within each of the three data sources, Selectome Euteleostomi [1000 amino-acid (AA) and nucleotide (NT) datasets each], Selectome Drosophila (1000 AA and NT datasets each), and PANDIT (236 AA and NT datasets each). Bars are colored based on the information theoretic criterion used for model selectionBack to article page