Drosophila Linkage Mapping Pipeline (DLinkMaP)
This R code was developed to conduct Quantitative Trait Loci (QTL) mapping based on haplotype probabilities for the Drosophila Synthetic Population Resource (DSPR). The code performs a linear mixed model analysis at 10kB intervals throughout the Drosophila genome. Based on the round robin crossing design described in Dew-Budd et al. (2018, in preparation), random effect terms are included for month, round robin group, RIL by month, and cross (all of these are included by default in the software). Using haplotype frequencies as found in the "DSPRqtlDataA", test founder probabilities corresponding to all 8 x 8 = 64 founder by maternal/paternal combinations are calculated. Six separate non-null models corresponding to additive, dominant, and full found by parent effects as main effects and diet interactions were included in the model (all of these models are included by default in the software). Statistical inference is conducted by calculating the χ2-statistics based on the difference in model log-likelihoods in a hierarchical manner, and determining p-values based on the χ2-distribution.
The software has the capability to conduct epistasis (gene-gene interaction) modeling by calculating the cross product of the haplotype frequencies for the additive model. Dominant and full models are not considered for epistatic modeling. By default the software calculates epistatic effects for all pairwise 10kB positions in the genome. As in the non-epistatic analysis, the statistical inference is based onχ2-statistics and negative log p-values are reported.
For each significant peak (by default the complete genome includes 11,768 QTLs, thus a negative log p-value above 5.37 is bonferroni adjusted significant), the range of influence for each peak, 95% confidence intervals is computed using a LOD drop of two. For the epistatic models, a marginal confidence interval was calculated for each QTL involved in a significant epistatic interaction (threshold of six for the –log p-value). A Bayesian model was conducted to estimate the variance explained for each non-epistatic peak and the significant epistatic interactions (more details are described in supplemental methods).
Running DLinkMaP: By running the "MAP_general.R" script with appropriate inputs (see below), the QTL mapping will be conducted (using the user defined phenotype) for all QTLs in DSPR Population A. The following R scripts are included in the pipeline:
MAP_general.R:
Conduct QTL Mapping for the user defined phenotype ('metabolites', 'triglyceride', 'male weight', 'female weight', and 'trehalose' have been tested)
Inputs:
commDir: directory where all QTL Mapping scripts are contained
p: Whether permutation testing should be (p=1) or should not be (p=0) conducted
outDir: directory where all output will be written
phenotype: which phenotype is being analyzed - 'metabolite', 'weight', 'trehalose', and 'TG' have been tested
weight.sex: if the 'weight' phenotype is being analyzed, which sex is analyzed - for all other phenotypes, ignore this!
weight.type: if the 'weight' phenptype is being anayzed, whether the original data or the average by vial is being analyzed - for all other phenotypes, ignore this!
fileName: the filepath and name of the dataset to be analyzed
epistaticModel: True/False if an epistatic model is run. F is normal mapping with additive, dominant and full effects; T is additive with epistatic effects for all pairs of QTLs
epistaticQTL: Integer (out of the total number of QTLs tested) for the epistatic model
MAPFun_general.R; FUN.R; gradMM.R; MM_Process.R; QC.R; QTL_Process.R These script contain key functions which are used by MAP_general.R to conduct the design matrix setup, PCA, and inference.
Necessary R packages include:
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent