guesses relations between individuals...

findRelatives {GenABEL}R Documentation

guesses relations between individuals...


guesses relations between individuals


findRelatives(gtdata, nmeivec=c(1:2), q, epsilon=0.01, quiet=FALSE,
    OddsVsNull=1000, OddsVsNextBest=100, twoWayPenalty=log(10),
    doTwoWay=TRUE, vsIDs, gkinCutOff, kinshipMatrix)


gtdata genotypic data, either '' or '' class, or matrix or 'databel' matrix (see details for format).
nmeivec vector providing the degree of relationship to be tested (1: parent-offspring; 2: sibs, grandparent-grandchild; etc.).
q vector of effect allele frequencis for the data.
epsilon genotyping error rate
quiet if TRUE, screen outputs supressed
OddsVsNull threshold used in relationships inferences (see details)
OddsVsNextBest threshold used in relationships inferences (see details)
twoWayPenalty penalty on likelihoods resulting from models assuming two meiotic pathways
doTwoWay or not
vsIDs specific IDs to be tested vs others
gkinCutOff if not null, sets a threshold used to pre-screen pairs before guessing relations. If value < 0 provided, procedure sets threshold automatically (recommended)
kinshipMatrix (genomic) kinship matrix (used if gkinCutOff!=NULL)


This function guesses relationships (expressed as estimated number meiotic connection(s)) using genomic data. Compared to guessing relations from genomic kinship matrix, this procedure offers several enhancements:

(1) by use of IBD/IBS 3-state space, it allows to distinguish between some pairs, which have the same kinship (e.g. parent-offspring from brother-sister; uncle-nephew from grandparent-grandchild, etc.)

(2) it reports likelihood, allowing for more rigorous inferences

If 'gtdata' are provided as a matrix (or 'databel' matrix), genotypes should be coded as 0, 1, or 2; each SNP corresponds to a column and each ID is a row. 'q' corresponds to the frequency of 'effect' (aka 'coded') allele, which is also equivalent to the mean(SNP)/2.0 provided coding is correct.

'nmeivec' is a sequence of integers, e.g. c(1,2) will test for parent-offspring pairs and pairs separated by two meioses (sibs, grandparent-grandchild, etc.). If 'nmeivec' does not contain '0' as its first element, it will be automatically added (testing for twins). Also, nmeivec will be updated with c(nmeivec,max(nmeivec)+1,100) to allow for thesting of testing vs. 'null' (unrelated, 100) and 'most distant' specified by user (max(nmeivec)).

While one may be interested to test only a sub-set of the data for relationships, it is recommended to provide 'q' estimated using all data available (see example).

'gkinCutOff' allows use of genomic kinship matrix (computed internally) to pre-screen pairs to be tested. Use of this option with value '-1' is recommended: in this case threshod is set to 0.5^(max(nmeivec)+2). If not NULL, only pairs passing gkinCutOff are tested with the likelihood procedure.

After likelihood estimation, inference on relationship is made. Releationship (in terms of number of meioses) is 'guessed' if odds of likelihoods under the meiotic distance providing max likelihood and under the 'null' (maximal meiotic distance tested + 100) is greater than 'OddsVsNull' parameter AND odds max-lik vs. the next-best meiotic distance is greater than 'OddsVsNextBest' parameter.


A list with elements call – details of the call; profile – table detailing likelihood for all pairs tested; estimatedNmeioses – nids x nids matrix containing maximum likelihood estimate of meiotic distance for all pairs of individuals guess – same as estimatedNmeioses, but all estimates not passing inference criteria (OddsVsNull, OddsVsNextBest) are NAed; compressedGuess – same as above, but removing cows and cols with missing-only elemnts


df <- ge03d2.clean[,autosomal(ge03d2.clean)]
df <- df[,sort(sample(1:nsnps(df),1000))]
eaf <- summary(gtdata(df))$"Q.2"
relInfo <- findRelatives(df[27:30,],q=eaf)
# look only for 1st and 2nd degree relatives
relInfo1 <- findRelatives(df[27:30],q=eaf,gkinCutOff=-1,nmeivec=c(1,2,3))
relInfoVS <- findRelatives(df[27:30,],q=eaf,nmeivec=c(1:6),vsIDs=idnames(df[27:30,])[1:2])

[Package GenABEL version 1.6-7 Index]