Problem 3: Simulated Rheumatoid Arthritis Data

Michael B Miller <mbmiller@taxa.epi.umn.edu>

Michael Li <nali@umn.edu>

Gregg Lind <lind1199@umn.edu>

Soon-Young Jang <jangsoon@taxa.epi.umn.edu>

The plan for this data simulation was to mimic the familial pattern of rheumatoid arthritis (RA) including a strong effect of DR type at the HLA locus on chr 6 and other genetic and environmental effects. For each of 100 replicates, we generated a large population of about two million nuclear families (two parents and two offspring) with RA affection status determined by a complex genetic/environmental model, then we retained a random sample of 1,500 families from those families that had an affected sibling pair (ASP) and a random sample of 2,000 families where none of the four members were affected (control). We present data on all members of the 1,500 ASP families and on one randomly-selected member of the offspring generation from each of the 2,000 unaffected control families (i.e., there are 2,000 unrelated control subjects per replicate, and no control subject had a first-degree relative with RA).

Here are some results of analysis of data from a simulated general population of 1,800,000 sibling pairs (3,600,000 subjects) generated using the model we developed:

RA lifetime prevalence 0.0107

F:M sex ratio in affecteds 3.07

Lambda_sib 9.03

Number of ASPs 1856 (of 1.8 million sib pairs)

"Lambda_sib" is the lifetime prevalence in siblings of affected individuals (probandwise concordance) divided by the lifetime prevalence in our simulated general population. The numbers in the table above are similar to what one would see in real RA epidemiological data. These numbers apply to the offspring generation, and not the parent generation of our simulated data.

We present markers on 22 autosomes which were designed to be like real human autosomes in terms of genetic and physical map lengths, but we did not generate data for sex chromosomes. The markers are in three sets:

1. A set of 730 microsatellite markers, fairly evenly spaced on chromosomes with an average inter-marker distance of about 5 cM and with heterozygosities always exceeding .70.

2. A set of 9,187 SNPs distributed on the genome to mimic a 10K SNP chip set but without monomorphic SNPs.

3. A very dense map of 17,820 SNPs on chromosome 6 (an average inter-marker spacing of 9,586 bp which corresponds roughly to the density one would expect from a genome-wide 300K SNP set). The chr 6 dense map includes 210 of the markers from the 10K SNP map (they are easily identifiable because they have the same names in both sets).

The data for marker and trait loci were generated so that they will have many of the properties of real data. For example, you should observe LD over short genetic distances both between pairs of markers and between markers and trait loci and the patterns of LD should be similar to patterns seen in real data.

Data Files - Filenames and Formats:

The data consist of both map information, with lists of markers and their locations, and simulated marker and phenotype data. The map data are stored in maps.tar.gz, and they are described below following the discussion of the simulated data which are stored in a different collection of files.

Simulated Marker and Phenotype Data Files

The marker and phenotype data from the 100 replicates are stored in gzipped tar files named rep0001.tar.gz, rep0002.tar.gz, ..., rep0100.tar.gz. Each of these compressed files uses about 79 MB of disk space, but when uncompressed and extracted they create directories named rep0001, rep0002, ..., rep0100, each of which uses about 891 MB of disk space. To uncompress and extract the files from the compressed archives, the following command seems to work on any unix system that has gunzip installed:

gunzip -c rep0001.tar.gz | tar xvf -

Simply replace 0001 with appropriate digits to extract other replicates. If you are using Linux, Cygwin, FreeBSD, Mac OS X, or some other Unix system, you can use this shorthand command to accomplish the same thing:

tar zxvf rep0001.tar.gz

If you are using Windows, we recommend that you either install Linux in a dual-boot configuration with Windows or run Cygwin under Windows (because of all the additional functionality those options provide for genetic analysis), but if those ideas are unappealing, the program 7-zip (http://www.7-zip.org/) should work for you. Unfortunately, 7-zip requires that uncompressing and extracting be done in two steps which takes up about 891 MB of extra disk space temporarily.

If you are using Mac OS 9, the data files can be extracted using Stuffit Expander (http://www.stuffit/mac/expander/).

The names of files within the replicate directories have the formats shown below where * stands for the replicate number padded with zeros (0001, 0002, ..., 0100), # stands for the chromosome number, ASP refers to the affected sibling pair nuclear families, CONTROL to the unrelated control subjects, SNP to SNP marker data, STRP to microsatellite data and dense.SNP to the dense SNP data from chromosome 6. The PHENOTYPE files contain all of the phenotype data for ASP families and control subjects.

Phenotype data filenames:

ASP.*.PHENOTYPE.ped

CONTROLS.*.PHENOTYPE.ped

Marker data filenames:

chr#.*.CONTROLS.SNP.ped

chr6.*.CONTROLS.dense.SNP.ped

chr#.*.ASP.SNP.ped

chr6.*.ASP.dense.SNP.ped

chr#.*.ASP.STRP.ped

Thus, for every replicate, there are 68 files of marker data (22 chromosomes times two types of subjects [ASP nuclear families and control subjects] for SNPs, 22 chromosomes of microsatellites for ASP families only and dense SNPs for chromosome 6 for both ASP families and controls) and two phenotype files: a total of 70 data files per replicate.

All fields (data columns) in every pedigree file (*.ped) are delimited by a single space. The first five columns always consist of these fields:

Family IDs are strings consisting of replicate number, underscore, family number (e.g., family ID 71_1367 corresponds to the 1,367th family in the 71st replicate). All individual IDs are unique integers within every replicate, but all replicates use the same sets of individual Ids. Control subject individual IDs (numbers 10001 to 12000) differ from ASP individual IDs (numbers 1 to 6000). This system makes it easy to mix together data from cases and controls from the same replicates or from different replicates. Following the first five columns are either phenotypes (in the *.PHENOTYPE.ped files) or markers. The markers always consist of two alleles per marker, separated by a single space (in other words, there are two space-delimited fields per marker). All alleles are integers from 1 to 20, but they are not necessarily consecutive integers because some alleles might not be observed. All SNP loci are diallelic and alleles are coded as 1 and 2.

Consecutive columns of the phenotype data (*.PHENOTYPE.ped files) are separated by a single space. After the first five columns described above, the remaining eleven columns consist of phenotypes in the following order:

6 Rheumatoid arthritis affection status (2=affected, 1=unaffected)

7 Dead (1=dead, 0=not dead)

8 Age at ascertainment (in years)

9 Lifetime smoking (1=smoked, 0=never smoked)

10 Anti-CCP continuous measure

11 IgM continuous measure

12 Severity (1 to 5; 1=mild, 5=severe)

13 DR allele from father

14 DR allele from mother

15 Age at death (missing if alive)

16 Age at onset (only available in offspring)

Map Data Files

The map data are stored in the file maps.tar.gz which expands to a directory named maps. The files within the maps directory have names with the following forms where # represents the chromosome number:

chr#.SNP.map

chr#.STRP.map

chr6.dense.SNP.map

The SNP and STRP in the filenames mean that the map data are for SNPs or for microsatellites (Simple Tandem Repeat Polymorphisms). Only chromosome 6 has a very dense SNP map. Within the .map files we report sex-averaged, male and female map locations (in Haldane cM) for all markers. We also present the physical location in base pairs. All fields in map files are separated by a single space and this is the format of every line of every map file:

<chr. no.> <marker name> <cM sex-ave.> <cM male> <cM female> <base pairs>

The tables below show the numbers of markers per chromosome for the three kinds of marker files:

microsatellite markers

count file

67 chr1.STRP.map

56 chr2.STRP.map

49 chr3.STRP.map

38 chr4.STRP.map

45 chr5.STRP.map

41 chr6.STRP.map

41 chr7.STRP.map

31 chr8.STRP.map

35 chr9.STRP.map

38 chr10.STRP.map

32 chr11.STRP.map

36 chr12.STRP.map

24 chr13.STRP.map

28 chr14.STRP.map

22 chr15.STRP.map

27 chr16.STRP.map

26 chr17.STRP.map

27 chr18.STRP.map

17 chr19.STRP.map

22 chr20.STRP.map

11 chr21.STRP.map

17 chr22.STRP.map

730 total

SNP markers

count file

704 chr1.SNP.map

813 chr2.SNP.map

687 chr3.SNP.map

642 chr4.SNP.map

622 chr5.SNP.map

674 chr6.SNP.map

479 chr7.SNP.map

442 chr8.SNP.map

475 chr9.SNP.map

472 chr10.SNP.map

492 chr11.SNP.map

496 chr12.SNP.map

406 chr13.SNP.map

334 chr14.SNP.map

257 chr15.SNP.map

204 chr16.SNP.map

156 chr17.SNP.map

303 chr18.SNP.map

93 chr19.SNP.map

187 chr20.SNP.map

174 chr21.SNP.map

75 chr22.SNP.map

9187 total

chr 6 dense SNP markers

count file

17820 chr6.dense.SNP.map

Remember that combining the two chromosome 6 maps gives us a total of 18,284 distinct SNP markers on chromosome 6.

In addition to the map files, we supply some files in a maps/MERLIN directory that use the MERLIN .dat format to specify the contents of the marker map and phenotype files.

Unusual Features of the Simulated Data:

We decided to provide more information in the simulated data than one would ordinarily have in real data. This allows the analyst to do some interesting things.

No Missing Data: One usually expect to be missing marker information on some family members, especially those who died before the family was ascertained. We provide marker data on all family members. Researchers who would like their data to be more realistic can delete marker information from deceased individuals. By supplying data that would normally be missing, we provide more opportunity to test effects of missingness, etc.

No Errors: We did not model any errors in the data simulation. In real data there are typically some errors in genotyping and sometimes there are sample mixups. By not modeling any errors, we make it possible for the analyst to simulate his own errors and test the effect of genotyping error on other aspects of a genetic analysis. We also added no errors to phenotypes.

Allele ordering in the output: The allele inherited from the father is always presented on the left side within every genotype. This allows researchers to determine haplotypes

for all subjects and to determine their parental origin. In real data, it is usually not possible to know haplotypes or their origin like this, but new methods have made molecular haplotyping possible and it is currently being used. So, in real data we can sometimes know haplotypes, but the parental origins of those haplotypes still must be inferred.

We hope you enjoy these data. We'll see you at the GAW meeting in November.

We thank the Minnesota Supercomputing Institute for their support for this simulation project. This work was supported, in part, by NIH grants 5RO1-HL09609-12, 1R01 AG021917-01A1, and by the University of Minnesota.