1 Introduction

The program RIA is a C++ implementation of the method described in Eu-Ahsunthornwattana et al. (2015) and uses calls to programs PLINK and KING version 2.2.9.

1.1 Overview

The program RIA performs the following steps calling external PLINK and KING where required:

  1. The first stage of RIA is the estimation of the IBD sharing probabilities between affected family members using a selection of SNPs across all available SNPs in the data, these probabilities form the priors. The program PLINK Purcell et al. (2007) is used to prune the SNPs to give a representative selection of SNPs using only founders. Thus, it should be noted that founders are required in the SNP data. The cases are then used to estimate the IBD sharing probabilities using KING (version 2.2.9) Manichaikul et al. (2010). See section 1.2 on how the IBD sharing probabilities are estimated.
  2. The next stage is to step across the genome, set to a default of 50 SNPs per step, and form a SNP window of around 500 to 2000 SNPs and use these SNPs to estimate the IBD sharing probabilities between affected family members using KING (version 2.2.9). These probabilities give the posteriors. See section 1.2 on how the IBD sharing probabilities are estimated.
  3. The next step is to perform a non-parametric linkage analysis comparing the prior and posterior IBD sharing probabilities using the method as described in Cordell et al. (2000) with minor modifications as described in Eu-Ahsunthornwattana et al. (2015). This produces a LOD score for each analysed SNP region together with parameter estimates for the (scaled) additive and dominance variances.

For full details of the methodology of RIA please see the accompanying paper Eu-Ahsunthornwattana et al. (2015).

1.2 Estimation of IBD sharing probabilities

KING version 2.2.9 is used with option --homog to estimate the IBD sharing probabilities. Let IBD2, IBD1 and IBD0 be the probabilities that two individuals share 2, 1 or 0 alleles IBD respectively and K the kinship coefficient, then IBD2, IBD1 and IBD0 are estimated for each SNP as follows:

  1. KING version 2.2.9 is ran with the --homog option which returns estimates of K and IBD0.
  2. K and IBD0 are truncated to plausible values between 0 to 0.5 and between 0 to 1 respectively if necessary.
  3. IBD2 is estimated as 4K - (1 - IBD0).
  4. IBD2 is truncated to a plausible value between 0 to 1 if necessary.
  5. IBD1 is estimated as 1 - IBD0 - IBD2.
  6. IBD1 is truncated to a plausible value between 0 to 1 if necessary.
  7. For the posterior probabilities only: if any of the priors for IBD2, IBD1 or IBD0 are equal to zero then the corresponding posterior values are also set to zero. If this results in all IBD probabilities being equal to zero then an error is reported.
  8. IBD2, IBD1 and IBD0 are scaled so that IBD2 + IBD1 + IBD0 = 1 if necessary.

next->