Skip to main content

Table 1 The number of datasets on which methods failed is indicated below by model condition

From: Statistically consistent divide-and-conquer pipelines for phylogeny estimation using NJMerge

# of taxa # of genes ILS level Sequence type Method # of failures (out of 20)
100 25 Very high Exon NJMerge(\({\mathcal {T}}_{true}\), \(D_{LD}\)) 1
100 25 Very high Exon NJMerge(\({\mathcal {T}}_{RAX}\), \(D_{AGID}\)) 1
100 25 Very high Intron NJMerge(\({\mathcal {T}}_{true}\),\( D_{AGID}\)) 1
1000 1000 Low/moderate Exon SVDquartets 20
1000 1000 Low/moderate Exon RAxML 3
1000 1000 Low/moderate Intron NJMerge(\({\mathcal {T}}_{AST}\), \(D_{LD}\)) 1
1000 1000 Low/moderate Intron SVDquartets 20
1000 1000 Low/moderate Intron RAxML 20
1000 1000 Very high Exon ASTRAL-III 19
1000 1000 Very high Exon NJMerge(\({\mathcal {T}}_{true}\), \(D_{LD}\)) 1
1000 1000 Very high Exon NJMerge(\({\mathcal {T}}_{AST}\), \(D_{LD}\)) 1
1000 1000 Very high Exon NJMerge(\({\mathcal {T}}_{SVD}\), \(D_{LD}\)) 2
1000 1000 Very high Exon NJMerge(\({\mathcal {T}}_{RAX}\), \(D_{LD}\)) 2
1000 1000 Very high Exon SVDquartets 20
1000 1000 Very high Intron ASTRAL-III 4
1000 1000 Very high Intron NJMerge(\({\mathcal {T}}_{SVD}, D_{LD}\)) 1
1000 1000 Very high Intron SVDquartets 20
1000 1000 Very high Intron RAxML 19
  1. ASTRAL-III failed due to running beyond the maximum wall clock time of 48 h; SVDquartets failed due to segmentation faults; RAxML failed due to running out of memory; NJMerge failed due to being unable to find a legal siblinghood. Note that NJMerge is described by the input set \({\mathcal {T}}\) of constraint trees and input distance matrix D; see “Performance study” section for more information on the notation