. 2005 Mar 1;2(1):39–66. doi: 10.1186/1479-7364-2-1-39

Table 1.

Description of unrelated haplotyping programs, divided into four classes based on method.

Program name	Algorithm	Output^a	Missing data^b	Assumptions^c	Key features	Limitations	MAX subjects, loci and type	Platform	Ref.^d
Parsimony methods
1. Simple parsimony
HAPAR	Parsimony	HA	No	None	Overcomes limitations of HAPINFREX	May be susceptible to HWE departures	Practical limit, biallelic	PC/UNIX	[39]
					Increasing sample size improves accuracy
HAPINFERX	Clark's	HA	No	None	Intuitive method, fast	May fail to start	Practical limit, biallelic/multiallelic	UNIX	[37]
					Reduced number of haplotypes	Sensitive to data order
					No limit on number of loci	Unstable and erroneous estimates
2. Phylogeny
BPPH	IP	HA	No	IP	Similar to HAP^H	User interface	Practical limit, biallelic	MAC	[45]
					Speed
DPPH	PP	HA	No	PP	Handles large datasets	Theoretical	Practical limit, biallelic	MAC	[40,43]
					Speed	Strict population assumptions
GPPH	PP	HA	No	PP	Handles large datasets	Theoretical	Practical limit, biallelic	MAC/PC/UNIX	[40,42]
					Speed	Strict population assumptions
HAP^H	IP	HA/HF	Yes	HWE, IP	Predicts haplotype blocks	No probability for haplotype assignments	Max 500 loci, Practical limit biallelic	Web-based	[44]
					Constructs haplotypes within blocks
					Identifies block structure
					Web-based
Likelihood methods
1. Maximum likelihood
Arlequin v2.0	EM	HA/HF	No	HWE	Includes numerous population genetic analysis tools	EM issues	EM Practical Limits, biallelic/multiallelic	JRE on MAC/PC/UNIX	[89]
CHAPLIN	ECM	HF	Yes	HWE	Graphical interface	ECM algorithm needs to be compared with standard EM methods	Practical limits, biallelic/multiallelic	PC	[91]
					Association tests
					HWE assumption relaxed in case sample
EH	EM	HF	No	HWE	Estimates haplotype frequency	EM issues	No Max, 3-4 practical max, biallelic/multiallelic	PC	[85,104]
					Compares case-control HF under different assumptions	Must specify mode of inheritance and penetrance of disease
EHPLUS	EM	HF	No	HWE	Improves EH, more loci and polymorphic markers	Long run times for permutation calculations	Max 5 loci, 15 alleles in analysis	PC/UNIX	[84]
					Incorporates model-free analysis
EM-DeCODER	EM	HA/HF	No	HWE	Program with standard EM algorithm	EM issues	Max 15 loci, biallelic	UNIX	[57]
FASTEHPLUS	EM	HF	No	HWE	Similar to EHPLUS, with speed improvements	EM issues	Max 5 loci, 15 alleles in analysis	PC/UNIX	[105]
GENECOUNTING	EM	HA/HF	Yes	HWE	Provides posterior probabilities for assigned haplotypes	Missing data limited to biallelic loci	10-15 loci practical limit, biallelic/ multiallelic	PC/UNIX	[106]
					Compares global and specific haplotypes between groups	EM issues
GCHAP	EM	HA/HF	YES	HWE	Haplotypes with zero likelihood dropped to improve speed and accuracy	EM issues	20 loci practical limit, biallelic	JRE on PC/UNIX	[107,108]
					Similar to SNPHAP
GS-EM	EM	HA/HF	Yes	HWE	Includes algorithm for assigning probability to genotype calls from several genotyping methods	EM issues	Practical limit, biallelic	Web-based	[73]
					Haplotypes constructed using assigned genotypes probability	Limited to biallelic SNPs
					Web-based
HAP^Z	EH	HA/HF	Yes	HWE	Modified version of SNPHAP that accommodates multiallelic loci	EM issues	Practical limit, biallelic/multiallelic	PC/UNIX	[106]
HAPMAX	MLE	HF	No	HWE	Ease of use	Accommodates a limited number of SNPs	8 loci, biallelic	PC	[109]
						Interface
HAPLO^H	EM	HF	Yes	HWE	Handles some missing data	EM issues	10 loci, 40 alleles max, biallelic/multiallelic	UNIX	[47]
					Utilises pedigree data, if available
					Calculates standard error
HAPLOSCOPE	EM/MCMC	†	†	†	Platform program, incorporates SNPHAP and PHASE v1.0	See individual programs for limitations/features	†	UNIX/Windows	[110]
					Facilitates comparison/testing
					Graphical interface, identifies tagging SNPs and LD blocks
HAPLOVIEW	EM+PL	HA/HF	Yes	HWE	Calculates pairwise LD	EM issues	100 s, practical limit, biallelic	JRE on MAC/PC/UNIX	[56]
					Checks for recombination
					Identifies tagging SNPs
					Accepts pedigree and unrelated genotype data
HAPLO.STATS	EM	HA/HF	Yes	HWE	Incorporates method similar to SNPHAP, with user inputs	Requires knowledge of S-Plus 6.0 or R	Practical limit, biallelic/ multiallelic	S-PLUS 6.0 on UNIX/R on UNIX & PC	[86]
					Separate programs that:	EM issues
					(1) assign haplotypes with posterior probability of assignments
					(2) allow linear regression for trait to haplotype analysis
					(3) calculates score statistic for haplotype phenotype association
HIT	EM/MCMC/ MC+PL	†	†	†	Platform program, incorporates SNPHAP and PHASE v1.0	See individual programs for limitations/ features	†	*	[111]
					Facilitates comparison
					Graphical interface, identifies tagging SNPs and LD blocks
HPLUS	EM+EE+PL	HA/HF	Yes	HWE	Provides posterior probabilities for assigned haplotypes	Requires Matlab	100 loci, biallelic	MATLAB on PC/ UNIX	[55,83]
					Compares haplotype frequencies between groups, adjusts for covariates	EM issues
					Utilises pedigree data, if available
LDSUPPORT	EM	HA/HF	Yes	HWE	Provides posterior probabilities for assigned haplotypes	EM issues	*	UNIX	[29,112]
					Identifies LD blocks for haplotype reconstruction
					Examines association with disease, automation speeds process
LOGINSERM ESTIHAPLO	EM	HA/HF	Yes	HWE	Program uses ML method to infer haplotypes for individuals with missing data	EM issues	Practical limit, biallelic/multiallelic	PC/ UNIX	[80]
					Offers option to exclude individuals with missing data
MLHAPFRE	EM	HF	Yes	HWE	Performance improves with presence of LD	Incorporated into Arlequin	16 loci, biallelic	JRE on Mac/PC/UNIX	[48]
					Performs well with large sample size	EM issues
MLOCUS	EM	HA/HF	Yes	HWE	Provides posterior probabilities for assigned haplotypes	EM issues	11 loci, biallelic/multiallelic	PC	[46,113]
					Notes observed vs. inferred haplotypes
					Calculates pairwise LD
OSLEM	EM	Yes	No	HWE	Modified EM algorithm that runs 2 × faster	EM issues	Practical limit, biallelic	Web-based	[114]
PL-EM	EM+PL	HA/HF	Yes	HWE	Combines PL with EM	EM issues	100 s, practical limit, biallelic	PC/UNIX	[54]
					EM-based version of HAPLOTYPER
					Calculates variance of haplotype frequency estimates
SAS Genetics	EM	HA/HF	Yes	HWE	Provides posterior probabilities for assigned haplotypes	Requires SAS	Practical limit, biallelic/multiallelic	SAS on PC/UNIX	[115]
					Incorporates statistical tests and procedures	EM issues
SNPEM	EM	HF	No	HWE	Estimates haplotype frequency by population	EM issues	10 loci, biallelic	UNIX	[10]
					Compares global and specific haplotype between 2 groups
SNPHAP	EM	HA/HF	Yes	HWE	Uses posterior and prior trimming to handle large number loci	EM issues	Practical limit, biallelic	UNIX	[52]
					Provides posterior probabilities for assigned haplotypes
THESIAS	S-EM	HF	Yes	HWE	Stochastic EM avoids issues of standard EM programs	S-EM algorithm needs to be compared with standard EM methods	Practical limit, 20 loci, biallelic	PC/UNIX	[53,88]
					Includes tests for haplotype-phenotype association
					Accommodates large sample sizes
WHAP	EM	†	†	†	Uses haplotype output from SNPHAP for association testing	EM issues	†	PC/UNIX	[116]
					Allows weighted association analysis	Requires separate haplotyping program
Zaykin et al.	EM	HF	No	HWE	Program on analysis of haplotype-phenotype association	EM issues	Practical limit, biallelic/multiallelic	PC/UNIX	[82]
						Subjects with missing data ignored
Zou and Zhao	MLE/EM	HF	Yes	HWE	Adjust haplotype frequency estimates for genotyping error	Assumes genotyping errors are random	Practical limits, biallelic/multiallelic	*	[68]
					Program also works for nuclear families	Assumes error rates are known
3locus.PAS	EM	HF	Yes	HWE	Handles some missing data	EM issues	3 loci, biallelic/ multiallelic	PC/UNIX	[46]
					Various tests available
					Improves with increasing sample size
2. Simple Bayesian
HAPLOTYPER	MC+PL	HA/HF	Yes	HWE	Uses PL algorithm to construct haplotypes with many loci	Long run times	256 max, biallelic	UNIX	[57]
					Provides posterior probabilities for assigned haplotypes	Posterior probabilities may be difficult to interpret
HAPLOREC	MC-VL	HA/HF	Yes	HWE	Uses variable length chain based on maximising LD	Restarts avoid non-global optimum	Practical limit, biallelic	Java virtual machine, v1.4 or newer	[62]
					Handles large number loci
3. Coalescent-based Bayesian^e
Arlequin v3.0	ELB	HA/HF	No	Adaptive window	Includes numerous population genetics analyses	Long run times	1,000 s, biallelic/ multiallelic	JRE on LINUX/ PC/Mac	[60,89]
					Handles recombination
PHASE v2.0	MCMC+PL	HA/HF	Yes	Coalescent/ HWE	Improved run time	Departure for coalescent model may impact performance	Practical limit, biallelic/ multiallelic	PC/MAC/ UNIX	[59]
					Compares haplotype frequency between groups	Posterior probabilities may be difficult to interpret
					Handles recombination
					Provides posterior probabilities for assigned haplotypes
PHASE v1.0	MCMC	HA/HF	No	Coalescent/ HWE	Incorporates pop-genetics and coalescence ideas	Departures for coalescent model may impact performance	Practical limit, biallelic/ multiallelic	UNIX	[51]
					Incorporates known phase and trios pedigrees into analysis	Slow run times
					Provides posterior probabilities for assigned haplotypes	Posterior probabilities may be difficult to interpret
SLHAP v1.0	MCMC	HA/HF	Yes	Neutral coalescent/ HWE	Similar to PHASE v1.0	Departures for coalescent model may impact performance	Practical limit, biallelic/multiallelic	UNIX	[58]
					Missing data
					Improved run time

^aProgram haplotype output, individual assignment, frequency estimates or both.

^bAbility of program to accept missing data.

^cProgram assumptions.

^dList of references.

^ePrograms in this section make assumptions based on or draw inference from coalescent model.

*Could not determine from available data.

^†See incorporated programs for features and limitations.

EE: Estimating equation; ECM: Expectation conditional maximisation algorithm; ELB: Excoffier-Laval-Balding algorithm, Bayesian; EM: Expectation maximisation algorithm; EM issues: May be sensitive to HWE departures, long run times, and non-global max (requiring multiple restarts); HF: Haplotype frequency estimate; HA: Individual haplotype assignment; HWE: Hardy-Weinberg equilibrium; IP: Imperfect phylogeny-based method; JRE: Java runtime environment; LD: Linkage disequilibrium; MAC: Program runs on Apple computer; MC: Monte Carlo algorithm, Bayesian algorithm; MCMC: Markov Chain Monte Carlo algorithm, Bayesian algorithm; MC-VL: Monte Carlo-variable length chain algorithm, Bayesian Algorithm; MLE: Maximum likelihood estimation algorithm; PC: IBM compatible personal computer; PL: Partition ligation algorithm; PP: Perfect phylogeny-based method; Practical Limit: program has no upper limit on number of markers and/or subjects, however computational and practical considerations limit this value; S-EM: Stochastic EM algorithm; UNIX: Runs on Unix operating system, including Linux, FORTRAN, Solaris and others.