Extended Data Fig. 6: Evolutionary lineage reconstruction without preliminary MSA.
From: Deep distributed computing to reconstruct extremely large lineage trees
a, Sample tree reconstruction and phylogenetic placement of FRACTAL with no preliminary multiple sequence alignment (MSA). A given number of sequences are first randomly subsampled from the input sequences (Step 1). The subsampled sequences are aligned with a common root sequence by MSA using MAFFT (Step 2) and a sample tree is reconstructed by a software tool of choice (Step 3). Each of the remaining input sequences are then independently added to the MSA result by ‘plus-one’ alignment using HMMER (Step 4i) and placed on the sample tree (Step 4ii). b, Accuracies of reconstructing various sizes of clades in the reference lineage of 1,000,000 sequences generated by RNASim. c, Accuracies and coverages of reconstructing the entire lineage of 1,000,000 unaligned sequences by FRACTALized RapidNJ, RAxML and FastTree with 100 computing nodes (five trials). d, Time series for the numbers of computing jobs and waiting jobs observed in the reconstruction of the simulated lineage of RNA evolution using 100 computing nodes for FRACTALization.
