Contents

In this vignette, we want to estimate the weight~covariate function on the RNASeq and eQTL datasets, without using the k-fold strategy. The below plots are mainly useful for presentations.

library("IHW")
library("dplyr")
library("ggplot2")
library("grid")
library("tidyr")
library("cowplot")
library("IHWpaper")
red_col <- scales::hue_pal(h = c(0, 360) + 15, c = 100, l = 65, h.start = 0,direction = 1)(1)

1 DESeq2 Bottomly

Load Bottomly dataset, apply DESeq2, then IHW with 1 fold.

bottomly <- analyze_dataset("bottomly")
##   Note: levels of factors in the design contain characters other than
##   letters, numbers, '_' and '.'. It is recommended (but not required) to use
##   only letters, numbers, and delimiters '_' or '.', as these are safe characters
##   for column names in R. [This is a message, not a warning or an error]
## estimating size factors
##   Note: levels of factors in the design contain characters other than
##   letters, numbers, '_' and '.'. It is recommended (but not required) to use
##   only letters, numbers, and delimiters '_' or '.', as these are safe characters
##   for column names in R. [This is a message, not a warning or an error]
## estimating dispersions
## gene-wise dispersion estimates
## mean-dispersion relationship
##   Note: levels of factors in the design contain characters other than
##   letters, numbers, '_' and '.'. It is recommended (but not required) to use
##   only letters, numbers, and delimiters '_' or '.', as these are safe characters
##   for column names in R. [This is a message, not a warning or an error]
## final dispersion estimates
## fitting model and testing
##   Note: levels of factors in the design contain characters other than
##   letters, numbers, '_' and '.'. It is recommended (but not required) to use
##   only letters, numbers, and delimiters '_' or '.', as these are safe characters
##   for column names in R. [This is a message, not a warning or an error]
## -- replacing outliers and refitting for 4 genes
## -- DESeq argument 'minReplicatesForReplace' = 7 
## -- original counts are preserved in counts(dds)
## estimating dispersions
## fitting model and testing
##   Note: levels of factors in the design contain characters other than
##   letters, numbers, '_' and '.'. It is recommended (but not required) to use
##   only letters, numbers, and delimiters '_' or '.', as these are safe characters
##   for column names in R. [This is a message, not a warning or an error]
ihw_res <- ihw(bottomly$pvalue, bottomly$baseMean, 0.1,
               nfolds=1, nbins=13, 
               nfolds_internal=4L, nsplits_internal=5L)
## Using only 1 fold! Only use this if you want to learn the weights, but NEVER for testing!

Do the plotting:

breaks <- IHW:::stratification_breaks(ihw_res)
break_min = min(bottomly$baseMean)

breaks_left <- c(break_min,breaks[-length(breaks)])

step_df <- data.frame(weight=weights(ihw_res,levels_only=TRUE),
                      stratum = 1:nlevels(groups_factor(ihw_res))) %>%
          mutate(baseMean_left = breaks_left[stratum], 
                 baseMean_right = breaks[stratum],
                 baseMean_ratio = baseMean_right/baseMean_left , 
                 baseMean_left = baseMean_left * baseMean_ratio^.2,
                 baseMean_right = baseMean_right *baseMean_ratio^(-.2))

stratum_fun <- function(df){
  stratum <- df$stratum
  weight <- df$weight
  stratum_left <- stratum[stratum != length(stratum)]
  weight_left  <- weight[stratum_left]
  baseMean_left <- df$baseMean_right[stratum_left]
  stratum_right <- stratum[stratum != 1]
  weight_right <- weight[stratum_right]
  baseMean_right <- df$baseMean_left[stratum_right]
  data.frame(stratum_left= stratum_left, weight_left= weight_left, 
             stratum_right = stratum_right, weight_right = weight_right,
             baseMean_left = baseMean_left, baseMean_right = baseMean_right)
}

connect_df <-  step_df %>% 
               do(stratum_fun(.)) %>%
               mutate(dashed = factor(ifelse(abs(weight_left - weight_right) > 10^(-4) , TRUE, FALSE),
                         levels=c(FALSE,TRUE)))


panel_b <- ggplot(step_df, 
                       aes(x=baseMean_left, 
                           xend=baseMean_right, 
                           y=weight, yend=weight)) +
                geom_segment(size=0.8, color=red_col)+ 
                geom_segment(data= connect_df, aes(x=baseMean_left, xend=baseMean_right, 
                                                y=weight_left, yend=weight_right, 
                                                linetype=dashed),
                                                size=0.8, color=red_col)+
                scale_x_log10(breaks=c(1,10,100,1000,10000))+
                xlab("Mean of normalized counts")+
                theme(legend.position=c(0.8,0.4)) +
                theme(plot.margin = unit(c(1, 1, 2, 1), "lines")) +
                guides(linetype=FALSE) + 
                theme(axis.title = element_text(face="bold") )

panel_b

ggsave(panel_b, filename="bottomly_1fold_weight_function.pdf", width=7, height=5)

2 hQTL example

Next we load the data for the hQTL example. Here we load the data-frame with the snps, genes and pvalue information. Note that because of memory concerns (the full data-frame is 3 GB) it contains only the p-values below 0.005. However, before doing this reduction we counted the number of hypotheses m_groups in each stratum (strata as defined in group column). Thus we can apply IHW to the reduced dataset, see also the advanced section of the IHW vignette (“Advanced usage: Working with incomplete p-value lists”).

hqtl_filt <- system.file("extdata/real_data",
                        "hqtl_pvalue_filtered.Rds", package = "IHWpaper")
hqtl_filt <- readRDS(hqtl_filt)
m_groups <- attr(hqtl_filt, "m_groups")

ihw_qtl_res <- ihw(hqtl_filt$pvalue, as.factor(hqtl_filt$group), 0.1,
                   m_groups = m_groups, nfolds=1L, lambda=Inf)
## Using only 1 fold! Only use this if you want to learn the weights, but NEVER for testing!
breaks <- attr(hqtl_filt, "breaks")
breaks <- breaks[-1]
break_min <- 5000
breaks_left <- c(break_min,breaks[-length(breaks)])
step_df_qtl <- data.frame(weight=weights(ihw_qtl_res,levels_only=TRUE),
                    stratum = 1:nlevels(groups_factor(ihw_qtl_res))) %>%
               mutate(break_left = breaks_left[stratum],
                    break_right = breaks[stratum],
                    break_ratio = break_right/break_left , 
                    break_left =break_left * break_ratio^.2,
                    break_right = break_right *break_ratio^(-.2))

stratum_fun <- function(df){
  stratum <- df$stratum
  weight <- df$weight
  stratum_left <- stratum[stratum != length(stratum)]
  weight_left  <- weight[stratum_left]
  break_left <- df$break_right[stratum_left]
  stratum_right <- stratum[stratum != 1]
  weight_right <- weight[stratum_right]
  break_right <- df$break_left[stratum_right]
  data.frame(stratum_left= stratum_left, weight_left= weight_left, 
             stratum_right = stratum_right, weight_right = weight_right,
             break_left = break_left, break_right = break_right)
}

connecting_df_qtl <- step_df_qtl %>% 
                  do(stratum_fun(.)) %>%
                  mutate(dashed = factor(ifelse(abs(weight_left - weight_right) > 5, 
                                      TRUE, FALSE),
                                  levels=c(FALSE,TRUE)))

panel_f <- ggplot(step_df_qtl, aes(x=break_left, xend=break_right,y=weight, yend=weight)) +
                geom_segment(size=0.8, color=red_col)+                
                geom_segment(data= connecting_df_qtl, aes(x=break_left, xend=break_right, 
                                                y=weight_left, yend=weight_right, 
                                                linetype=dashed),
                             size=0.8, color=red_col) +
                scale_x_log10(breaks=c(10^4, 10^5,10^6,10^7)) +
                xlab("Genomic distance (bp)")+
                theme(legend.position=c(0.8,0.4)) +
                theme(plot.margin = unit(c(1, 1, 2, 1), "lines"))+
                guides(linetype=FALSE) + 
                theme(axis.title = element_text(face="bold") )

panel_f

ggsave(panel_f, filename="hqtl_1fold_weight_function.pdf", width=7, height=5)