Table 1.

Feature comparisons between QMaker and two previously published estimation procedures (Whelan and Goldman 2001; Le and Gascuel 2008)

Feature	Whelan and Goldman	Le and Gascuel	QMaker
Tree reconstruction	Neighbor-joining (Saitou and Nei 1987) with DayhoffF distances	PhyML (Guindon and Gascuel 2003) with WAG4 model	IQ-TREE(Nguyen et al. 2015) with best-fit model
Branch length estimation	Scaled on JTTF	ML estimate	ML estimate
Rate heterogeneity across sites	No	Gamma model (Yang 1994)	Gamma(Yang 1994), invariant site (Gu et al. 1995), and free rate model (Kalyaanamoorthy et al. 2017)
Algorithm to optimize the ML parameters	Expectationmaximization	Expectation maximization	Broyden–Fletcher–Goldfarb–Shanno (Fletcher 1987)
Multicore CPU support	No	No	Yes
Explicit separation of training and testing data	No	No	Yes