Table 2. Key properties of imputation methods.

Name	Can Use Reference Panel	Can Perform Phasing	Can Use Genotype Likelihood Inputs	Requires Prephasing	Can use Prephasing to Speed up Computation	Computational Complexity
BEAGLE v.4.1a (Browning and Browning 2016)	Yes	No	No	Yes	Yes	Linear in number of haplotypes
BEAGLE v.4.0 (Browning and Browning 2007, 2009)	Yes	Yes	Yes	No	No	Quadratic in number of haplotypes
fastPHASE (Scheet and Stephens 2006)	Yes	Yes	No	No	No	M-step linear in number of haplotypes, quadratic in number of clusters
IMPUTE (v.1) (Marchini et al. 2007)	Yes	No	No	No	No	Quadratic in number of haplotypes
IMPUTE2 (v.2) (Howie et al. 2009, 2011)	Yes	Yes	Yes	No	Yes	Phasing quadratic in number of haplotypes, imputing linear in number of haplotypes
MaCH (Li et al. 2010)	Yes	Yes	No	No	No	Quadratic in number of haplotypes
MiniMac (Howie et al. 2012)	Yes	No	No	Yes	Yes	Linear in number of haplotypes (phasing precomputed)
MiniMac2 (Fuchsberger et al. 2015)	Yes	No	No	Yes	Yes	NA
SNPTools (Wang et al. 2013)	No	Yes	Yes	No	No	Constant in number of haplotypes
GeneImp	Yes	No (but can be added)	Yes	No	No	Subquadratic in number of reference haplotypesb

^aAttributes for the options that invoke the v.4.1 algorithm in the BEAGLE software package.

^bGeneImp hidden state considers reference haplotypes only (i.e., ignores haplotypes of other target individuals).