LCA: Model Averaged Line Cross Analysis

In coleoguy/SAGA2: Software for the Analysis of Genetic Architecture

Model Averaged Line Cross Analysis

Description

Analyze all possible genetic architecture models based on mean phenotypes from line cross data.

Usage

LCA(data, 
    SCS = "XY", 
    parental = "calc", 
    env = FALSE, 
    model.sum = .95,
    max.models = 300000, 
    drop.pars = NULL, 
    keep.pars = NULL,
    max.pars = NULL, 
    Cmatrix = NULL,
    ret.all = FALSE,
    messages = TRUE,
    Mepi = FALSE)

Arguments

data	a data frame with the 7 columns: cross: the name of each cohort e.g. P1, F1, BC1 etc.3 mean: phenotype measure of the cohort SE: standard error of the cohort's mean phenotype sex: sex of the cohort "M", "F", "E", and "U" representing male, female, equal ratio, unknown or unequal ratio are valid enviro: a numeric value indicating an environmental variable measured for each cohort - can be NA or NULL sire: the row of the user data that represents the sire of the current line dam: the row of the user data that represents the dam of the current line
SCS	A text string used to indicate the sex chromosome system "XY", "XO", "ZW", "ZO", or "NSC" (No Sex Chromosomes) are all valid.
parental	"calc" or "obs" "calc": a traditional calculated genetic maternal effect will be used "obs": an observed phenotypic parental effect (maternal and paternal) will be used - see details for more information.
env	TRUE or FALSE indicating whether to include E and GxE interactions.
model.sum	This is the sum of the probability of the models to be included
max.models	The maximum number of fitted models to return from the function. This is included as an option to allow analysis of large model space on computers with limited RAM.
drop.pars	Vector of strings with the names of composite genetic effects to ignore.
keep.pars	Vector of strings with the names of composite genetic effects to test.
max.pars	Optional numeric value indicating the maximum number of parameters for equations evaluated.
Cmatrix	Optional matrix describing contribution of genetic effects in the cohorts being analyzed.
ret.all	logical indicating if results of fitting all models should be returned
messages	logical indicating if verbose messages should be returned
Mepi	logical indicating if interactions between maternal effects and standard genetic effects should be allowed - has not been tested.

Details

Provides model averaged estimates of the contribution of composite genetic effects to the line means in line cross analysis experiments. Using AICc models are given weights and these are used to construct a confidence model set that allows for parameter estimates and errors to include model selection uncertainty. (see Burnham and Anderson 2002). The original paper contains a full discussion of the implementation (Blackmon and Demuth 2016).

In previous LCA work maternal effects have always been modeled as the autosomal additive or dominance status of the dam. To the degree that the phenotype of interest is not controlled by either autosomal additive or dominance effects this approach will fail to capture the true nature of maternal effects. As an alternative we offer the parental argument. If supplied with the value "obs" The phenotypes of all cohorts will be rescaled on an interval of -1 to 1. These values will then function as the expected contribution of either maternal or paternal effects in offspring from a cohort.

Example Data Set

cohort	mean	SE	sex	enviro	sire	dam
P1	32	.45	E	32	1	1
P2	16	.65	E	32	2	2
F1	24	1.2	E	32	1	2
rF1	27	3.1	E	32	2	1
BC1a	24	1.2	E	32	1	3
P1	29	.01	E	29	6	6
P2	12	.98	E	29	7	7
F1	22	.8	E	29	6	7
rF1	21	2.4	E	29	7	6
rBC1a	20	.56	E	29	8	6

• Cohorts can be named in any way the user whishes but P1 and P2 are standard names that must be used for the two original strains. If these are not present in the user data then the software will not be able to construct a C-matrix.

• Note that in the case of BC1a the dam is indicated to be a mix of both rF1 and F1 females. The software will assume this is an equal mixture and is not able to analyze data with unequally pooled lines.

• This dataset also indicates that crosses were conducted in two environments (29 and 32). These will be rescaled appropraitely and treated as a potential environmental or gene by environmental effect on the phenotype being studied.

Value

Returns a "genarch" object which is a list with the following elements:

all.models:	a list containing the weighted least squares solution for all models tested.
best.models:	a list containing the weighted least squares solution for all models in the confidence set.
best.eqns.w:	a vector containing the model probabilities for each model in the confidence set.
estimates:	a data frame containing Model Weighted Average for each parameter and its unconditional standard error.
daicc:	a vector of the delta AICc scores for all models tested.
varimp:	a data frame containing the variable importance scores for composite effects.
cmatrix:	a data frame containing the contribution of possible genetic effects used in the analysis.

Author(s)

Andrew Armstrong, Nathan Anderson, and Heath Blackmon

References

Burnham, K. P., & Anderson, D. R. (2002). Model selection and multimodel inference: a practical information-theoretic approach. Springer.

Blackmon, H. and Demuth, J.P., (2016). An information-theoretic approach to estimating the composite genetic effects contributing to variation among generation means: Moving beyond the joint-scaling test for line cross analysis. Evolution, 70(2), pp.420-432.

coleoguy/SAGA2 documentation built on Feb. 2, 2023, 2:15 p.m.