Pedigree and genotype errors in the Framingham Heart Study
© Brush and Almasy; licensee BioMed Central Ltd 2003
Published: 31 December 2003
The pedigree and genotype data from the Framingham Heart Study were examined for errors. Errors in 21 of 329 pedigrees were detected with the program PREST, and of these the errors in 16 pedigrees were resolved. Genotyping errors were then detected with SIMWALK2. Five Mendelian errors were found following the pedigree corrections. Double-recombinant errors were more common, with 142 being detected at mistyping probabilities of 0.25 or greater.
Because linkage analysis observes the co-segregation of marker alleles and phenotype, there is a concern that errors in pedigrees or genotypes could result in false-negative or false-positive results. The use of Genetic Analysis Workshop (GAW) data to examine the presence and nature of pedigree errors is not new [1–3]. In this exercise we detect and describe the pedigree and genotyping errors in the GAW13 Framingham Heart Study data.
Pedigree error detection and correction
The program PREST  was used to detect pedigree errors. PREST estimates the probabilities, p0, p1, and p2 of two individuals sharing 0, 1, and 2 alleles identically by descent (IBD), respectively. We calculated this over all of the relationship pairs known to PREST (parent-offspring, full-sibs, half-sibs, avuncular, first-cousins, grandparent-grandchild, half-avuncular, half-first cousin, half-sib plus first-cousin, monozygotic twins, and unrelated) within and between pedigrees. Pedigree errors were first screened with PREST's analytical tests: conditional estimated identity by descent (EIBD), adjusted identity by state (AIBS), and IBS, in that order and where applicable, at α = 0.0001, to focus on the more significant problems. This index pair and their relatives were then examined more thoroughly using PREST's accompanying program ALTERTEST that can test two individuals for each of the 11 relationship classes.
PREST comes with an R script written by Dan Weeks to plot the IBD estimation of a single relative pair on a relationship triangle . We modified this program to provide a scatter diagram of IBDs on the triangle. The result is an informative graphical summary of the pedigree errors in the sample. Pedigrees were drawn with PEDIGREE/DRAW .
Genotyping error detection and correction
Genotyping errors are detected using SIMWALK [7, 8]. SIMWALK2 applies a Markov-chain Monte Carlo method to data from the pedigree, population allele frequencies, and a genetic map to assign probabilities of mistyping for each genotype. Because this is a computationally intensive exercise, we examined genotyping errors only on chromosome 7.
We ran SIMWALK2 in two phases. In the first phase, Mendelian errors were detected and corrected independently for each marker. Marker genotypes were blanked (changed to a missing value) for all probabilities of mistyping above a given threshold. The threshold was chosen conservatively, i.e., to blank no more genotypes for a marker than necessary to calculate a likelihood for that marker. The mistyping probability was decremented from 1.0 until a calculable likelihood was reached.
In the second phase, genotypes that suggest improbable double recombination events were blanked. Mistyping probabilities were assigned using the genetic maps supplied with the GAW13 data. In this phase, the proportion of genotypes potentially blanked at a series of thresholds is plotted to provide a visual guide for choosing a threshold.
These simple examples belie the difficulties of correcting pedigree errors. In the Framingham data we were able to resolve the errors in 16 of 21 pedigrees, out of 329 pedigrees in total. The five unresolved pedigrees displayed patterns of errors that contradicted all testable alternative hypotheses, and were left unchanged.
Following the blanking, a new genetic map for the 21 markers on chromosome 7 was estimated using MULTIMAP/CRIMAP [9, 10]. The new map was 167 cM in length, or 24 cM shorter than the map provided with the data. In comparison, the length of the corresponding Marshfield map is 175 cM .
The Framingham data were relatively free of pedigree errors, particularly those involving close relatives. The most frequent type of error involved individuals in the upper generations, detected through their descendants due to the lack of genotype data for the ancestors. These errors were frequently corrected through the joining of disconnected pedigrees. It is difficult to generalize these findings to other populations in which the types and distribution of pedigree errors may be quite different. Presumably in Framingham, kinship terms are usually given a biological interpretation. In other populations where this is not strictly the case, serious errors would seem possible.
The small number of Mendelian genotyping errors in the Framingham data that had not been detected previously became evident only following the pedigree corrections. Genotype errors that imply unlikely double recombination events were more common, the exact number depending on a definition of probable mistyping.
We are grateful to Linda Freeman-Shade (SFBR) for providing PEDSYS  versions of PREST (PREPREST) and SIMWALK2 (PRESWALK), thus making these programs very easy to use, and Thomas Dyer (SFBR) for support programs for use with SIMWALK2. This research was partially funded by NIH grant MH59490.
- Broman KW: Cleaning genotype data. Genet Epidemiol. 1999, 17 (suppl 1): S79-S83.View ArticlePubMedGoogle Scholar
- Sun L, Abney M, McPeek MS: Detection of mis-specified relationships in inbred and outbred pedigrees. Genet Epidemiol. 2001, 21 (suppl 1): S36-S41.PubMedGoogle Scholar
- Cherny SS, Abecasis GR, Cookson WOC, Sham PC, Cardon LR: The effect of genotype and pedigree error on linkage analysis: analysis of three asthma genome scans. Genet Epidemiol. 2001, 21 (suppl 1): S117-S122.PubMedGoogle Scholar
- Sun L, Wilder K, McPeek MS: Enhanced pedigree error detection. Hum Hered. 2002, 54: 99-110. 10.1159/000067666.View ArticlePubMedGoogle Scholar
- Ihaka R, Gentleman R: A language for data analysis and graphics. J Comp Graph Stat. 1996, 5: 299-314. 10.2307/1390807.Google Scholar
- Mamalka PM, Dyke B, MacCluer JW: Pedigree/Draw for the Apple Macintosh, Technical Report No. 1. San Antonio, TX, Population Genetics Laboratory, Southwest Foundation for Biomedical Research. 1989Google Scholar
- Sobel E, Lange K: Descent graphs in pedigree analysis: applications to haplotyping, location scores, and marker-sharing statistics. Am J Hum Genet. 1996, 58: 1323-1337.PubMed CentralPubMedGoogle Scholar
- Sobel E, Papp JC, Lange K: Detection and integration of genotyping errors in statistical genetics. Am J Hum Genet. 2002, 70: 496-508. 10.1086/338920.PubMed CentralView ArticlePubMedGoogle Scholar
- Matise TC, Perlin M, Chakravarti A: Automated construction of genetic linkage maps using an expert system (MultiMap): a human genome linkage map. Nat Genet. 1994, 6: 384-390. 10.1038/ng0494-384.View ArticlePubMedGoogle Scholar
- Green P, Falls K, Crooks S: CRI-MAP Documentation. St. Louis, MO, Department of Genetics, Washington University in St. Louis. 1990Google Scholar
- Broman KW, Murray JC, Sheffield VC, White RL, Weber JL: Comprehensive human genetic maps: individual and sex-specific variation in recombination. Am J Hum Genet. 1998, 63: 861-689. 10.1086/302011.PubMed CentralView ArticlePubMedGoogle Scholar
- Dyke B: PEDSYS, a Pedigree Data Management System User's Manual, Technical Report No. 2. San Antonio, TX, Population Genetics Laboratory, Southwest Foundation for Biomedical Research. 1994Google Scholar
This article is published under license to BioMed Central Ltd. This is an open access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.