BXD Family

Download datasets and supplementary data files.
BXD genotype current revision 050423
BXD genotype 2017
BXD genotype 2001-2016

(Updated July 1, 2022, by D. Ashbrook)

All variants are publicly available for anyone to get whatever type and frequency of variant that they want. The variant vcf is under analyses files in project PRJEB45429 https://www.ebi.ac.uk/ena/browser/view/PRJEB45429?show=analyses

(Updated March 15, 2018, by RW Williams)

BXD Genotypes file status (January 2017): From  September 2016 to January 2017, Robert Williams, Jesse Ingels, Lu Lu, and Danny Arends released a new genotype file for the original  BXD strains (BXD1 through BXD102)  and for all of the new strains (BXD104 to BXD220). Version 1 of this genotype file (used from Jan 2017 to March 13, 2018) contained data for 7324 markers and 198 strains. Version 2 of March 14, 2018, fixed some errors in marker location detected by Karl Broman (five markers were out of order in the latest mouse genome assembly). We deleted three markers and retained a final set of 7321 markers, now all in the correct order based on the SNP position using the mm10 assembly.

Of the 198 BXD strains, 191 are independent, whereas 7 are substrains (e.g., BXD48 and BXD48a). The file provides approximate locations of 10300 recombinations, an average of 52 per strain. Genotypes were generated using Affymetrix, MUGA, MegaMUGA, and GigaMUGA Illumina platforms. Microsatellites and eQTL genotypes were generated by the Williams/Lu laboratory. Unknown genotypes were imputed as B or D or were called H (heterozygous) if the genotype was uncertain. Genotypes were manually curated by RW Williams. Genotypes were smoothed to remove unlikely recombination events. Almost all recombinations are supported by multiple markers, although only one or two representative markers may be provided in this file. The original parent file (BXD_El_Grande_Master_Used_to_Proof_Final_Genotypes_2016.xlxs) contains data for approximately 37000 markers.   Genotypes for Chr Y and  Chr M are provisional and will be verified in 2017. As of 2016, many strains with higher numbers (BXD100 and above) are not fully inbred.

A link to the genotype file is provided here

Genotypes were generated at GeneSeek (Neogen Inc) with financial support from the University of Tennessee Center for Integrative and Translational Genomics. We thank  Drs. Fernando Pardo-Manuel de Villena (University of North Carolina) and Gary Churchill (The Jackson Laboratory) for developing the GigaMUGA array.

The new genotypes are now available in GeneNetwork as the 2017 Genotype file. All SNPs were mapped to the newer Dec 2011, mm10, GRCm38 assembly. 

As of Jan 2017, GeneNetwork uses mm10 coordinates for mapping functions.  Older mm9 versions of GeneNetwork are available on the GN TimeMachine (see upper right side of the Search page).

BXD Genotype: The state of a gene or DNA sequence, usually used to describe a contrast between two or more states, such as that between the normal state (wildtype) and a mutant state (mutation) or between the alleles inherited from two parents. All species that are included in GeneNetwork are diploid (derived from two parents) and have two copies of most genes (genes located on the X and Y chromosomes are exceptions). As a result, the genotype of a particular diploid individual is actually a pair of genotypes, one from each parent. For example, the offspring of a mating between strain A and strain B will have one copy of the A genotype and one copy of the B genotype and therefore have an A/B genotype. In contrast, offspring of a mating between a female strain A and a male strain A will inherit only A genotypes and have an A/A genotype.

Genotypes can be measured or inferred in many different ways, even by visual inspection of animals (e.g. as Gregor Mendel did long before DNA was discovered). But now the typical method is to directly test DNA that has a well define chromosomal location that has been obtained from one or usually many cases using molecular tests that often rely on polymerase chain reaction steps and sequence analysis. Each case is genotyped at many chromosomal locations (loci, markers, or genes). The entire collection of genotypes (as many as 1 million for a single case) is also sometimes referred to as the cases genotype, but the word "genome type" might be more appropriate to highlight the fact that we are now dealing with a set of genotypes spanning the entire genome (all chromosomes) of the case.

For gene mapping purposes, genotypes are often translated from letter codes (A/A, A/B, and B/B) to simple numerical codes that are more suitable for computation. A/A might be represented by the value -1, A/B by the value 0, and B/B by the value +1. This recoding makes it easy to determine if there is a statistically significant correlation between genotypes across a set of cases (for example, an F2 population or a Genetic Reference Panel) and a variable phenotype measured in the same population. A sufficiently high correlation between genotypes and phenotypes is referred to as a quantitative trait locus (QTL). If the correlation is almost perfect (r > 0.9) then the correlation is usually referred to as a Mendelian locus. Despite the fact that we use the term "correlation" in the preceding sentences, the genotype is actually the cause of the phenotype. More precisely, variation in the genotypes of individuals in the sample population causes variation in the phenotype. The statistical confidence of this assertion of causality is often estimated using LOD and LRS scores and permutation methods. If the LOD score is above 10, then we can be extremely confident that we have located a genetic cause of variation in the phenotype. While the location is defined usually with a precision ranging from 10 million to 100 thousand base pairs (the locus), the individual sequence variant that is responsible may be quite difficult to extract. Think of this in terms of police work: we may know the neighborhood where the suspect lives, and we may have clues as to identity and habits, but we still may have a large list of suspects.

The BXD genotype file was initially upgraded in 2010-2011 using the new high-density Affymetrix array (580,000 high-quality SNPs) developed in the laboratories of Drs. Fernando Pardo-Manuel de Villena (University of North Carolina) and Gary Churchill (The Jackson Laboratory, see Yang H, Ding Y, Hutchins LN, Szatkiewicz J, Bell TA, Paigen BJ, Graber JH, Pardo-Manuel de Villena, F, Churchill GA (2009) A customized and versatile high-density genotyping array for the mouse. Nat Methods 6:663-666)

The BXD genotype file used from June 2005 through December 2016 exploits a set of approximately 3796 markers typed across 88 extant and extinct BXD strains (BXD1 through BXD102). The mean interval between informative markers is about 0.7 Mb. This genotype file includes all markers, both SNPs and microsatellites, with unique strain distribution patterns (SDPs), as well as pairs of markers for those SDPs represented by two or more markers. In those situations where three or more markers had the same SDP, we retained only the most proximal and distal markers in the genotype file. This particular file has also been smoothed to eliminate genotypes that are likely to be erroneous. We have also conservatively imputed a small number of missing genotypes (usually over very short intervals). Smoothing genotypes is this way reduces the total number of SDPs and also lowers the rate of false discovery. However, this procedure also may eliminate some genuine SDPs.

The new smoothed BXD genotype data file (2023) can be downloaded from
BXD genotype current revision 050423.

Please Note: For a limited number of markers and strains, the genotypes of BXDs have been called heterozygous. This is usually done over comparatively short intervals in some of the newer strains that may not have been fully inbred when they were initially genotyped. Use of the genotype file above in external software packages such as R/qtl, requires careful treatment of this issue to prevent bias in empirical significance thresholds. It is recommended to treat these rare heterozygous loci as missing data and ensure that only the additive effects of B vs. D alleles are estimated by these packages. (note by Elissa Chesler, Dec 2010).

Source of Genotypes:

In collaboration with members of the CTC (Richard Mott, Jonathan Flint, and colleagues), we have helped genotype a total of 480 strains using a panel of 13,377 SNPs. These SNPs were combined with our previous microsatellite genotypes to produce the older "classic" consensus maps for the expanded set of BXD using the older mouse assemblies (Mouse Build 36 - UCSC mm8 and then mm9).  (Files were updated from mm6 to mm8 in January 2007, and from mm9 to mm10 in January 2017).

A total of 198 strains have been genotyped as of Jan 2017 using the full set of SNPs, and about 7324 of these are informative. Informative in this sense means that the C57BL/6J and DBA/2J parental strains have different alleles. To reduce false positive errors when mapping using this ultra-dense map, we have eliminated most single genotypes that generate double-recombinant haplotypes that are most commonly produced by typing errors ("smoothed" genotypes). For this reason, the genotypes used in the GeneNetwork differ from those downloaded directly from Richard Mott's website at the Wellcome Trust, Oxford, or from the Jackson Laboratory.

The BXD family of recombinant inbred (RI) strains were derived by crossing C57BL/6J (B6) and DBA/2J (D2) and inbreeding progeny for 20 or more generations. This genetic reference panel is a remarkable resource because data for thousands of phenotypes and nearly 100 gene, protein, and metabolite expression data sets have been acquired over a nearly a 40-year period. Another advantage of the BXD family is that the both parents have been sequenced (C57BL/6J as part of a public effort, and DBA/2J by Celera Genomics, by the UTHSC group and by Sanger). Based on our analysis of the sequence data, these two strains differ at approximately 4.8 million SNPs. Variants (mostly single nucleotide polymorphisms and about 500,000 insertion-deletions) that produce interesting phenotypes can be located efficiently. The zoomable physical maps in GeneNetwork can display the positions of B versus D-type SNPs at high resolution.

Our DBA/2J sequence data (from Wang et al. 2016) have been used to generate a virtual genome for this strain using a C57BL/6J framework. In other words, all SNPs and small DBA/2J indels have inserted in place of original C57BL/6J sequence. 

 

BXD strain

Legend: Photo gallery of BXD strains

EPOCH DIFFERENCES or "Batch Effect" among BXD strains. BXD strains (1 through 103) were produced as at least four separate groups or subfamilies. BXD1 through BXD30 were produced by Benjamin A. Taylor starting in about 1971, with the first publication using early generation BXD lines at F7 to F10 in 1973 (Taylor et al., 1973  Full text, 1975 (Taylor et al., 1975, Womack et al., 1975). A distinction is made between an RI line, which is not necessarily fully inbred (<20 F generations of inbreeding, and an RI strain, which should be the progeny of 20 or more sequential sib matings). 

BXD31 and BXD32 are exceptional BXD strains. They were created by the Mouse Mutant Resource at The Jackson Laboratory in the propagation of different visible mutations. Mutations arose in C57BL/6J or DBA/2 and then outcrossed to the other strain and subsequently inbred while maintaining the mutation. In the case of BXD31, inbreeding was followed after the F2 generation. In BXD32, there was a backcross to DBA/2 before sib mating. BXD32 is therefore approximately 75% DBA/2 and 25% B6. When Taylor learned about these strains he decided that they would be a useful addition to the BXD RI set. He eliminated the visible mutations, continued inbreeding, and added them to the BXDs set. The first publication using these two strains was in 1980. Both are usually lumped together with BXD1 through BXD30 as a "single" first cohort, but obviously, they should perhaps be excluded from all cohorts or BXD "epochs." BXD32 is also exceptional because it has a D mitochondrial genome and a B-type Y chromosome. (Information in this paragraph mainly from BA Taylor, email of Aug 17, 2014 to RWW).

BXD33 through BXD42 were also produced by Benjamin Taylor (Taylor et al. 1999), but from a new set of F2 crosses initiated in the early 1990s.

BXD43 through BXD102 were produced by Lu Lu, Jeremy Peirce, Lee M. Silver, and Robert W. Williams in the late 1990s and early 2000s using advanced intercross progeny (Peirce et al. 2004). These strains have roughly twice the number of recombinations as conventional F2-derived RI strains.

BXD104 through BXD157 were produced by Lu Lu and Robert W. Williams starting in 2008 using standard F2 stock. 

BXD160 through BXD186 were produced by Lu Lu and Robert W. Williams starting in 2010 using G8 and G9 advanced intercross stock donated by Abraham Palmer. 

BXD187 through BXD220 were produced by Lu Lu and Robert W. Williams starting in 2014 using F2 stock. 

Initial genotypes in 2008 highlighted breeding errors that resulted in sets of very closely related sister substrains that are almost genetically identical. In collaboration with the Jackson Laboratory, we have made the following changes in strain names to clarify the strong genetic relations among the newer BXD strains. 

In general, data for the following sister substrains needs to be handled with special care prior to or during statistical analysis or gene mapping for the simple reason that the substrains are not genetically independent. However, if investigators do discover significant phenotype differences among strains, then data can be treated independently. (The latest version of GeneNetwork (GN 2) includes a mapping method called Python Linear Mixed Model (pyLMM) that corrects for the effects of shared pedigrees and corrects mapping results (code written by Nick Furlotte in 2011-2013). )

  1. BXD73, BXD73a (original known as BXD80), and BXD73b (originally known as BXD103) are genetically very similar. BXD73 and BXD80 are genetically identical at 82264 of 100290 markers (82% identical by descent). BXD73 keeps its original name and JAX identifier number (JR#7117), whereas BXD80 is now referred to as BXD73a (JR#7124). BXD73 and BXD103 are genetically identical at 90917 of 100290 markers (90.6% identical by descent). BXD103 is now referred to as BXD73b (JR#7146). 
  2. BXD48 and BXD48a (originally known as BXD96) are sister substrains, and are genetically identical at 93485 of 100290 markers (93.2% identical by descent). BXD48 retains its original name and JAX identifier number (JAX JR#7097) whereas BXD96 is now referred to as BXD48a (JR#7139). 
  3. BXD65, BXD65a (originally known as BXD97), and BXD65b (originally known as BXD92) are sister substrains. BXD65 and BXD97 are genetically identical at 92225 of 100290 markers (92% identical by descent). BXD97 is now referred to as BXD65a (JR#7140). BXD65 and BXD92 are genetically identical at 6155 of 6459 markers (95.3% identical by descent). BXD65 retains its original name and JAX identifier (JR#7110) whereas BXD92 is now referred to as BXD65b (JR#9677).

 

While the strains used to generate these subsets of BXDs have the same official names and were all made using stock from the Jackson Laboratory, the individual parents were are not genetically identical due to inevitable genetic drift and mutation. Shifman and colleagues detected a surprisingly large number of new SNPs (n = 47 out of about 13000 SNPs studied) in the set of strains generated by BA Taylor in the early 1990s, and a small number (n = 5) of even newer SNPs in the set of BXD strains generated at UTHSC in the late 1990s (see Shifman et al., 2006).

"In the BXD set, 52 SNPs showed variation in genotypes that corresponded to the different phases of development of the BXD RIs [24–26] (Table S4). Forty-seven SNPs are not polymorphic in the 26 BXD strains established from a single cross of a C57BL/6J female to a DBA/2J male, but are polymorphic in similar BXD strains established more than 20 y later. Five SNPs are not polymorphic in the first 36 BXD strains, but are polymorphic in the newest set of 53 BXD lines (BXD43–100)."

Correction for Family or Epoch Substructure

The BXDs have the following epoch substructure:

  1. BXD1 through 30 make up the first epoch. Breeding for this group of BXD strains started in about 1970, with the first publication of fully inbred BXD strains in 1975 (Taylor et al., 1983, see Trait ID 10715). In fact, BXD32 has a mitochondrion that is inherited from DBA/2J. BXD32 could be considered the first DXB strain (DXB32).
  2. BXD33 to BXD42 make up Ben Taylor's final addition to the BXD strains (Taylor et al, 2001, Trait ID 10645).
  3. BXD43 to BXD103. This is a complex cohort of strains generated at UTHSC from advanced intercross progeny (Peirce et al., 2004). 
  4. BXD104 to BXD157. This is a single cohort of strains generated at UTHSC from F2 intercross stock. 
  5. BXD160 to BXD186. This is a single cohort of strains generated at UTHSC from G8 and G9 advanced intercross progeny donated to RW Williams by Abraham Palmer in 2008. 
  6. BXD187 to BXD220. This is a single cohort of strains generated at UTHSC from F2 intercross stock. 
Users of the expanded BXD panel should take this epoch substructure into account. This is easy to do using the "Epoch" traits that are included in the BXD Phenotype database. For example, BXD Phenotype 12688 (BXD epoch batch trait 1) provides a simple code for the major phases of BXD production using the code of -1 for the first set through to BXD32, 0 for the second set (33 to 42), and +1 for the newer UTHSC set (43 to 103). 
  1. Determine whether your trait covaries well with any one of the three Epoch traits in GeneNetwork. Also check the status of BXD31 and BXD32. They may not belong to any group.
  2. Determine if your trait maps extremely well to Chr 4 at 62 Mb (near the ALAD segmental duplication in DBA/2J).

Strain nomenclature: Some of the BXD strains have accumulated new mutations that have recently been characterized. When these mutations are known, the full nomenclature of the strain is now being modified. For example, BXD24/TyJ (aka BXD24 in most GeneNetwork databases), suffered a mutation in the Cep290 gene in the late 1980s. The mutant allele (rd16 is associated with autosomal recessive retinal degeneration. The original BXD strain was briefly referred to as BXD24a/TyJ, while the blind co-isogenic mutant was referred to as BXD24b/TyJ. The great majority of phenotype, expression, and genotype data in GeneNetwork was generated using these blind BXD24b/TyJ animals. However, in 2010, the nomenclature was changed again and the blind variant (JAX stock 000031) is now known as BXD24/TyJ-Cep290rd16/J. The original BXD with normal vision was rederived from frozen stock and is now known once again as BXD24/TyJ, although the stock number has now been changed to 005243.

BXD29/TyJ was also known as BXD29/TyJ-Tlr4, but is now formally BXD29-Tlr4lps-2J/J (JAX stock 000029). The original non-mutant stock is currently known as BXD29/TyJ again but the stock number of these rederived non-mutants has been changed to 010981.

The mitochondrial DNA of all BXD strains were typed by Jing Gu and Shuhua Qi (Nov 2004) using DNAs obtained from the Jackson Laboratory (BXD1 through 42) or from the UTHSC colony. This typing relied on a SNP marker identified by Jan Jiao in Weikuan Gu's laboratory at nucleotide position 9461 in the reference C57BL/6J mitochondrial sequence. Most strains have inherited mitochondria from C57BL/6J. However, the following strains have mitochondria with a Dallele at the rs8281487 (UT-M-9461) SNP in ND3 (T is reference and C is alternate allele): BXD32, 61, 74, 78, 82, 89, 90, 91, 95, and BXD99. These ten strains could be considered DXB recombinant inbred strains. 

Genotypes of these strains: All BXD strains were genotyped in the first half of 2005 at 13377 markers as part of a CTC-Wellcome Trust collaboration. When combined with previous markers, there are a total of 7636 informative markers that differ betweeen the parental strains and that are useful for mapping with the BXD strains. The locations of these makers are known on the latest assembly of the mouse genome (Build 34, mm6). The median distance between these informative markers is 178,831 bp. The mean distance is 324,493 bp. There are only 26 intervals between markers that are longer than 5 Mb. No interval is greater than 10 Mb except on Chr X. These long intervals are essentially monomorphic between the parental strains. 

The BXD genotype file used in GeneNetwork up to 2014 includes a selected subset of approximately 3795 markers (out of 7636) and includes all those markers with unique strain distribution patterns (SDP) as well as pairs of markers--the most proximal and most distal--for SDPs represented by two or more markers. Slightly updated versions of this BXD genotype data set can be downloaded by ftp at ftp://atlas.uthsc.edu/Public/BXD_WebQTL_Genotypes.

There are a total of 1848 known recombinations in the 36 older (JAX) BXD set; an average of 48.1 recombinations per strain. 

There are a total of 4366 known recombinations in the 53 of the first set of UTHSC BXD strains (BXD43 to BXD102); an average of 82.4 recombinations per strain (Shifman et al., 2006). These RI strains were generated from an advanced intercross, and this accounts for the higher recombination load (Peirce et al., 2005).

The "classic" genotypes of the BXD strains (used through Dec 2016 in GeneNetwork) rely mainly on the Mouse Universal Genotyping Array (MUGA) genotyping platform. They also rely to some extent and earlier genotyping in 2008 using the Affymetrix Mouse Diversity array and even earlier work by Williams et al (2001) using microsatellite markers.

In January 2017, the genotypes of most extant BXD strains were updated. The new data also include initial genotypes for the newest cohort of BXDs (BXD104 to BXD220). This January 2017 genotype file provides consensus genotypes for 198 BXD strains. Of the 198 BXD strains, 191 are independent, whereas 7 are substrains (e.g., BXD48 and BXD48a). This file provides approximate locations of 11500 recombinations, an average of 58 per strain. Genotypes were generated using Affymetrix, MUGA, MegaMUGA, and GigaMUGA Illumina platforms. Microsatellites and eQTL genotypes were generated by the Williams and Lu laboratory. Unknown genotypes were imputed as B or D, or were called as H (heterozygous) if the genotype was uncertain. Genotypes were manually curated by RW Williams. Genotypes were smoothed to remove unlikely recombination events. Almost all recombinations are supported by multiple markers, although only one or two representative markers may be provided in this file. The original parent file (BXD_El_Grande_Master_Used_to_Proof_Final_Genotypes_2016.xlxs) contains data for approximately 37000 markers. A subset of the most informative 7300 markers are included in the 2017 genotype file. Genotypes for Chr Y and Chr M are provisional and will be verified in 2017. As of 2016, many strains with higher numbers (BXD100 and above) are not fully inbred.

Approximately 5100 phenotypes are currently included in the BXD Phenotype database in GeneNetwork as of Jan 2017. You can get an update on this number by typing in an asterisk (*) in the search box. 

How to obtain these strains: Please see http://jaxmice.jax.org/strain/000105.html. Cost of the JAX BXD strains was approximately $65.40 each in 2008, $135.00 each in 2014, and $139.90 in 2017. To obtain strains BXD43 and higher please contact Rob Williams. All strains through BXD102 are now fully inbred. We expect to generate as many as 160 viable BXDs strains by 2020. All extant BXD strains are being sequenced in early 2017 at 30X using 10X Chromium libraries at Hudson Alpha (supported by the UTHSC CITG, Williams, Lu, and colleagues at UTHSC, Abraham Palmer and colleagues at UCSD, and by Jonathan Pritchard at Stanford).

For more details on the history, generation, and use of RI strains as genetic reference populations for systems genetics please see Silver (1995). Additional useful literature links are provided in the References link at the top center of this page.

We have genotyped all available BXD strains from The Jackson Laboratory. BXD1 through BXD32 was produced by Benjamin Taylor starting in the late 1970s. BXD33 through BXD42 was produced by Taylor in the 1990s (Taylor et al., 1999). All BXD strains with numbers higher than BXD42 (BXD43 through BXD100) were generated by Lu Lu and Robert Williams at UTHSC, and by Jeremy Peirce and Lee Silver at Princeton University. We thank Guomin Zhou for generating the advanced intercross stock used to produce most of these advanced RI strains both at UTHSC and Princeton. There are approximately 48 of these advanced BXD strains, each of which archives approximately twice the recombinations present in a typical F2-derived recombinant inbred strain (Peirce et al. 2003).

Mapping Algorithm:

Due to the very high density of markers, the mapping algorithm used to map BXD data sets has been modified and is a mixture of simple marker regression, linear interpolation, and standard Haley-Knott interval mapping. When two adjacent markers have identical SDPs, they will have identical linkage statistics, as will the entire interval between these two markers (assuming complete and error-free haplotype data for all strains). On a physical map, the LRS and the additive effect values will therefore be constant over this physical interval. Between neighboring markers that have different SDPs and that are separated by 1 cM or more, we use a conventional interval mapping method (Haley-Knott) combined with a Haldane estimate of genetic distance. When the interval is less than 1 cM, we simply interpolate linearly between markers based on a physical scale between those markers. The result of this mixture mapping algorithm is a linkage map of a trait that has an unusual profile that is particularly striking on a physical (Mb) scale, with many plateaus, abrupt linear transitions between plateaus, and a few regions with the standard graceful curves typical of interval maps.

Archival Genotypes:

Archival BXD Genotype file: Prior to July 2005, the marker genotypes used to map all BXD data sets consisted of a set of 779 markers described by Williams and colleagues (2001) that also included a small number of additional SNPs from Tim Wiltshire and Mathew Pletcher (GNF, La Jolla), new microsatellite markers generated by Grant Morahan and Jing Gu (Msw type markers), and a few CTC markers by Jing Gu. This old marker data set was made obsolete by the ultra-high density Illumina SNP genotype data generated in Spring, 2005.

Download Genotypes:

The entire BXD genotype data set used for mapping traits can be downloaded here: BXD genotype current revision 050423.

The majority of SNP genotypes were generated at GeneSeek using the GigaMUGA array, at UNC using the Affymetrix mouse genotyping array, and at Illumina with support from the Wellcome Trust. The selection of markers to be included in the final file was carried out by Robert W. Williams and Danny Arends in December 2017.

Reference:

Dietrich WF, Katz H, Lincoln SE (1992) A genetic map of the mouse suitable for typing in intraspecific crosses. Genetics 131:423-447.

Taylor BA, Wnek C, Kotlus BS, Roemer N, MacTaggart T, Phillips SJ (1999) Genotyping new BXD recombinant inbred mouse strains and comparison of BXD and consensus maps. Mamm Genome 10:335-348.

Williams RW, Gu J, Qi S, Lu L (2001) The genetic structure of recombinant inbred mice: High-resolution consensus maps for complex trait analysis. Genome Biology 2:RESEARCH0046

Wiltshire T, Pletcher MT, Batalov S, Barnes SW, Tarantino LM, Cooke MP, Wu H, Smylie K, Santrosyan A, Copeland NG, Jenkins NA, Kalush F, Mural RJ, Glynne RJ, Kay SA, Adams MD, Fletcher CF (2003) Genome-wide single-nucleotide polymorphism analysis defines haplotype patterns in mouse. Proc Natl Acad Sci USA 100:3380-3385.