Beyond classification: gene-family phylogenies from shotgun metagenomic reads enable accurate community analysis

BMC Genomics

Table 2 Minimum and maximum DFs for simulations with 200 reads from the rpoB gene family

Method	RefDB size	Mean read length	Mean min. DF	Std. dev. min. DF	Mean max. DF	Std. dev. max. DF
FastTree	Small	100 bp	3.8e-04	1.4e-04	6.9e + 02	8.5e + 02
	Small	400 bp	8.1e-04	4.0e-04	5.3e + 02	3.9e + 02
	Large	100 bp	3.9e-04	1.4e-04	4.6e + 02	3.4e + 02
	Large	400 bp	1.1e-03	7.0e-04	2.0e + 02	1.8e + 02
RAxML	Small	100 bp	2.2e-06	7.8e-07	4.6e + 05	3.4e + 05
	Small	400 bp	5.7e-06	3.3e-06	3.8e + 05	2.3e + 05
	Large	100 bp	2.3e-06	7.2e-07	1.2e + 05	1.5e + 05
	Large	400 bp	7.8e-06	4.2e-06	1.5e + 05	1.9e + 05
Pplacer	Small	100 bp	2.3e-06	5.5e-06	5.4e + 05	4.1e + 05
	Small	400 bp	3.4e-05	3.8e-05	4.7e + 05	2.4e + 05
	Large	100 bp	4.5e-06	7.6e-06	1.5e + 05	1.7e + 05
	Large	400 bp	2.3e-05	4.0e-05	2.0e + 05	2.0e + 05

Minimum and maximum DFs were orders of magnitude away from the ideal value of 1.0, indicating that the most poorly estimated branch lengths in each simulation were typically very inaccurate. Values were somewhat smaller for simulations with 50 reads (data not shown). The mean and standard deviation values of the minimum and maximum DFs are over 30 simulations for each parameter combination. Similar trends were observed for the other gene families.

ISSN: 1471-2164