Benchmarking software tools for detecting and quantifying selection in evolve and resequencing studies

Table 1 Overview of the evaluated tools

Tool	t	RAM	ts.	rep.	m/w	Description	Input	Output	lang.	Reference
χ ²	6 s	221 M	No	No	+/+	Pearson χ² test for homogeneity (vectorized implementation)	freq, cov, Ne	p	R	[14]
E&R- χ²	8 s	306 M	Yes	No	+/+	χ² test adapted to account for drift	freq, cov, Ne	p	R	[12]
CLEAR	3000 s	1100 M	Yes	Yes	+/+	Discrete HMM of allele trajectories under a WF model	sync,Ne	s, Ne, h, LL	Python	[11]
cmh	216 s	145 M	No	Yes	+/+	Test for homogeneity (similar to χ²) accounting for stratified data	sync	p	Perl/R	[13]
E&R-cmh	8 s	560 M	Yes	Yes	+/+	CMH test adapted to account for drift	freq, cov, Ne	p	R	[12]
LLS	1091 s (83 h)	340 M	Yes	Yes	+/+	Linear model with least square regression of logit-transformed allele frequencies	freq, cov, Ne	p, s, h	R	[14]
LRT-1	31 s	127 M	No	Yes	−/−	LRT of parallel selection	freq, cov, Ne	LRT, \(\hat \delta \)	Python	[15]
LRT-2	31 s	127 M	No	Yes	−/−	LRT of heterogeneous selection	freq, cov, Ne	LRT, dx_r	Python	[15]
GLM	220 s	300 M	Yes	Yes	+/+	Quasibinomial GLM with replicates and time as predictors	freq	p	R	[16]
LM	157 s	300 M	Yes	Yes	+/+	LM with replicates and time as predictors	freq	p	R	[16]
BBGP	37 h	15 M	Yes	Yes	+/+	A Bayesian model of allele trajectories following a Gaussian process	sync	BF	R	[17]
FIT1	16 s	220 M	Yes	No	−/−	A t test with allele trajectories modeled as a Brownian process	freq	p	R	[18]
FIT2	68 s	220 M	No	Yes	−/−	A t test with allele frequencies differences between two time points	freq	p	R	[18]
WFABC	42 h	8 MB	Yes	No	+/+	ABC of WF dynamics with selection	freq, Ne (h)	BF, s	C++	[20]
slattice	41 h	250 M	Yes	No	+/+	HMM of allele trajectories under a WF model using an EM algorithm	freq, Ne (h)	s, LL	R	[19]

For each tool, we show the time required to analyze a small data set (t, either in seconds (s) or hours (h)), the memory requirements (RAM), if time series data may be used (ts.), if replicates are accepted (rep), if a manual and a walk-through is available (m/w), a short description, the required input, the generated output, the programming language (lang.), and the reference for LLS the time required to estimate the selection coefficient and the p-value (in brackets) is provided. sync file, freq allele frequency, cov coverage, Ne effective population size, h heterozygous effect, p value, s selection coefficient, LRT likelihood ratio test, BF Bayes factor, LL log-likelihood, \(\hat \delta \) shared allele frequency change, dx_r change in allele frequency in a single replicate r

ISSN: 1474-760X