Fig. 2

Cassiopeia algorithms outperform other phylogenetic reconstruction methods on simulated lineages. Accuracy is compared between five algorithms (Cassiopeia-Greedy, Cassiopeia-ILP, and Cassiopeia-Hybrid algorithms as well as neighbor joining and Camin-Sokal) on 400 cells. Phylogeny reconstruction accuracy is assessed with the triplets correct statistic across several experimental regimes: a the number of characters, b the mutation rate (i.e., Cas9 cutting rate), c the depth of the tree (or length of the experiment), d the number of states per character (i.e., number of possible indel outcomes), and e the dropout rate. Dashed lines represent the default value for each stress test. Between 10 and 50 replicate trees were reconstructed, depending on the stability of triplets correct statistic and overall runtime. Standard error over replicates is represented by the shaded area