In byandell/qtl2shiny: Shiny interface for 'qtl2' Fine Mapping

knitr::opts_chunk$set(echo = FALSE)

shine a light on MAGIC data

Shiny tool to study DO and other MAGIC crosses

• What is MAGIC?
• Allele vs SNP Scans
• Additive Model is Quick & Easy
• Gene/SNP Action
• Shiny under the Hood

what is MAGIC?

• Multiparent Advanced Generation Inter-Cross (MAGIC)
  • experimental populations with >2 segregating alleles
• advanced generations yield many meiotic events
  • typically low linkage disequilibrium
  • capable of fine mapping in one pass
• de Koning DJ, McIntyre LM (G3 2014)
  • animals: mouse, Drosophila
  • plants: maize, wheat, rice, sorghum, Arabidopsis, Pigeonpea
  • mapping populations: AIL, CC, HS, DO, DSPR

Attie/Jax DO population

• 8 CC founder strains (generation 19-21)
• 400 mice in 4 waves
• multiple traits measured
  • clinical traits (insulin secretion)
  • 150K SNPs, 30K RNA-Seq
  • proteomic, metabolomic, lipidomic

---

allele vs SNP scans

• allele-based genome scan: LOD maps
  • continuous curve across loci
  • interval mapping for missing data
  • model effect of founder alleles
• DO founder alleles: A,B,C,D,E,F,G,H
• response ~ sum of effects of alleles
• additive model
  • $y$ ~ $a_1 + a_2$ (1st & 2nd allele)
  • test if all a's are the same
  • 8 unknown parameters

simple story on chr 1

---

SNP-based genome scan

• SNP-based genome scan: GWAS Manhattan plots
  • discrete tests of SNPs or other features
  • typically 2 SNP alleles (1 is reference)
  • model effect of number of non-ref SNP copies
• SNP recorded as pair of DNA base pairs (A,C,G,T)
  • SNPs typically have two values (G/T)
  • individual has genotype GG, GT or TT
• simplifed to number of copies of non-reference allele
  • $s$ = 0,1,2
  • DO reference is B = B6
• additive model:
  • $y$ ~ $a + bs$ (2 unknowns)
  • test slope: $b=0$?

---

---

additive model is quick & easy

• good news
  • fewer parameters to understand
  • easy to build tools
  • fast to run
  • relate LOD to allele effects directly
  • likely good model for mRNA expression
• bad news
  • could miss important gene action information
  • could miss loci (false negatives)
  • could detect false positives
  • LOD and allele effects may conflict due to dominance

full model (with dominance)

• DO founder alleles: A,B,C,D,E,F,G,H
• response ~ effect of pair of alleles
• full model for allele-based scan
  • $y$ ~ $a_1 + a_2 + d_{12}$ (additive & dominance effects)
  • $y$ ~ $\texttt{mean}(A_1,A_2)$ (alleles $A_1$, $A_2$)
  • test if all means are the same (all $a$s equal, all $d$s zero)
  • 36 unknown parameters
• full model for SNP-based scan
  • $y$ ~ $\texttt{mean}(s)$
  • test if all means are the same
  • 3 unknowns

gene/SNP action

• study additive & dominance of single trait
• compare co-mapping traits -- same gene action?
• extend discrete SNP-based scan to continuous
  • scan of particular allele contrasts
  • LOD and allele contrast scans

data <- data.frame(x=rep(0:2,4),y=c(0:2,0,1.5,2,0,2,2,0,0,2),
                   z=rep(c("additive","general","dominant","recessive"), each = 3))
library(ggplot2)
ggplot(data, aes(x,y)) +
  geom_point(size=4) +
  geom_line(size=2) +
  facet_grid(~z) +
  labs(x="",y="phenotype mean") + 
  theme(axis.text.y = element_blank()) +
  scale_x_continuous(breaks=0:2)

complicated story on chr 3

---

---

additive & dominance together

general SNP contrasts

• identify 3-level SNP with strong LOD
• interpret as contrast involving 36 allele pairs
  • example AB:CDEFGH
• divide subjects into 3 groups at each locus
  • AA,AB,BB allele pairs (3)
  • CC,CD,...,GH,HH allele pairs (21)
  • rest allele from AB and allele from CDEFGH (12)
• scan region for this set of contrasts
  • LOD scan with 2 df of AB:het:CDEFGH
  • allele group scan

additive BFGH:ACDE

B=B6 recessive, D=NOD dominant

B=B6 dominant, G=PWK recessive

shiny under the hood

• work flow
• tools and resources
• challenges remaining

lambda architecture for doqtl2

• batch layer (DOQTL)
  • genotypes
  • phenotype LOD scans across genome
  • phenotype LOD & allele scans for chromosome
  • phenotype SNPs & genes in peak region
• serving layer
  • web serving layer
• speed layer
  • Shiny \"speed\" layer
  • phenotype LOD & allele scan for chromosome
  • detailed look at peak region

tools & resources

• Original batch pipeline: DOQTL (Dan Gatti et al. 2014 G3)
• R/qtl2 (Karl Broman) (Karl Broman et al. in progress)
  • qtl2geno, qtl2scan, qtl2plot
  • web site
• new R packages
  • R/qtl2 package suite (Karl Broman)
  • R/doqtl2 package
  • R/qtl2shiny package
• Derived Data in various file forms
  • SQLite for mouse SNPs and gene features
  • RDS R data objects for genotypes
  • CSV comma separated variable for phenotypes

challenges remaining

• Shiny user interface from Rstudio
  • translates from R to HTML via javascript
  • run under Rstudio or via Shiny server
  • R/qtl2shiny has 20+ shiny modules
  • translate to Python?
• analysis & computation
  • calibration of significance thresholds
  • dimension reduction (filtering) for massive phenotypes
  • incorporation of causal network tools
• operation & connectivity
  • connecting to other omic resources
  • fast management of raw, derived and intermediate data
• visualizations
  • scalable & interactive (D3) visual displays
  • automated report generation

data storage issues

• raw data
• derived data
  • 3Gb gene, SNP & SVS features (mouse)
  • 2Gb haplotype probabilities
  • 10Gb diplotype probabilities
  • 0.2Mb clinical phenotypes
  • 1-2Gb molecular phenotypes
  • spatial/image phenotypes?
• intermediate data
  • tables & summaries
  • plot data (or saved plots)
  • on-the-fly vs pre-stored vs as-needed
• portability issues
  • CSV vs RDS vs SQL vs ...
  • common pool vs on-site

software issues

• R analysis & visualization
  • QTL tools: qtl2 squite: geno, scan, plot
  • discovery: doqtl2, qtl2shiny
  • interactive: shiny, shinydashboard
  • wrangle data: dplyr, tidyr, readr, stringr
  • graphics: ggplot2, grid
• phenotype & genotype pipelines
  • variety of languages and formats
  • in domain of biologists (or chemists or ...)
• high volume pipelines
  • hadoop/map-reduce technology
  • high throughput phenotypes
  • data resampling for thresholds
• scaling up to muli-user interactive system

byandell/qtl2shiny documentation built on June 11, 2025, 4:54 a.m.