About

In GAIT2, we recruited 935 individuals grouped in 35 pedigrees, with on average of 27 individuals per pedigree and a total of 8654 related pairs. All families have at least 10 living individuals in three or more generations, and were recruited and selected as part of a study on idiopathic thrombophilia with an age range of 3 to 101 years (mean 40). Hundreds of quantitative phenotypes were measured at time of recruitment, including anthropometric measurements, hemogram, hemostasis traits, as well as phenotypes related with platelets, platelet activity, homocysteine metabolism, inflammation, and flow cytometry of leukocytes and microparticles. Genotypes obtained from a mix of Illumina 1M and Illumina 500K chips were imputed into a 1000 genomes reference panel. We also sequenced the mRNA transcriptome from whole blood for all the individuals.

QTL association mapping

Around a hundred traits were mapped using around 10 millions of SNPs. The ultra-fast score-based GLS method MatrixeQTL allowed to efficiently evaluate the association models.

• Interactive Heatmap traits vs. SNPs, MAF >= 0.01, genome-wide significant level < 5e-8
  • Static Heatmap traits vs. SNPs, MAF >= 0.01, genome-wide suggestive level < 1e-5
• Interactive Scatterplot h2r vs. #SNPs, MAF >= 0.01

The results are stored in the private repository of the group GAIT2, directory projects/02-assoc-mapping2-matrix. Some links of the interest (all to the private group repository):

• Manhattan plots produced for all traits under study: link to download PDF.
• Directory output/assoc/ contains CSV tables of the association results for all the traits.
• Report reports/01-comparison-F11/01-comparison-F11.md presents a comparison of GWAS results produced by LMM method and score-based GLS method (used here for QTL asssociation mapping).

MatrixeQTL under GAIT2-specific settings seems to show consistent results in terms of the genome-wide significant signals, but the overall tendency is towards overestimation of the p-values. Thus, the reported signals are needed to be further refined with a more precise approach. In general, the prioritization of SNPs produced by MatrixeQTL is a good starting point to cope with the large-scale problem of QTL mapping.