kompute

Overview

The kompute R package is specifically designed to handle the imputation of missing summary statistics in high-throughput model organism data, such as the International Mouse Phenotyping Consortium (IMPC) dataset. Despite the IMPC’s extensive gene-phenotype cataloging efforts, exploring the entirety of the mouse genome remains a formidable task due to the volume of potential gene-phenotype pairs and the constraints of resource and time-intensive testing in a controlled environment. Certain gene knockouts can lead to early embryonic lethality or developmental abnormalities, rendering numerous phenotypes unmeasurable. Consequently, the IMPC dataset, though rich, is still considerably incomplete. kompute addresses this by leveraging the conditional distribution properties of multivariate normal data, allowing it to impute missing phenotype association statistics directly. The method employs only the summary statistics of measured phenotypes and the corresponding estimated phenotype correlations as proxies for genetic correlations between phenotypes.

Citing kompute

Warkentin et al. KOMPUTE: Imputing summary statistics of missing phenotypes in high-throughput model organism data. Bioinformatics Advances. doi: https://doi.org/10.1093/bioadv/vbad100

Authors

Donghyung Lee leed13@miamioh.edu and Coby Warkentin warkenca@miamioh.edu

Installation

To install the kompute package, start R and enter the following commands:

# install.packages("devtools")
devtools::install_github("dleelab/kompute")

Examples

Below is a basic example illustrating the usage of the kompute() function to impute missing Z-scores in IMPC Body Composition summary statistics data (BC_Zscore_Mat). This example utilizes estimated phenotype correlations (BC_Pheno_Cor) as proxies for genetic correlations between the Body Composition phenotypes.

library(kompute) # all example datasets will be automatically loaded
library(ComplexHeatmap)

#str(BC_Zscore_Mat) # association z-score matrix of Body Composition data
#str(BC_Pheno_Cor)  # Body Composition phenotype correlation matrix

BC_Zmat <- BC_Zscore_Mat[1:500,] # we only use the first 500 genes

# Calculate the percentage of missing Z-scores in the data 
pct_missing <- 100*sum(is.na(BC_Zmat))/(nrow(BC_Zmat)*ncol(BC_Zmat)) 
pct_missing
#> [1] 4.975

Visualization of Phenotype-Gene Coverage

The heatmap illustrates tested (blue) and untested (light gray) gene-phenotype pairs.

# Generate a matrix indicating where z-scores are present
id.mat <- 1*(!is.na(BC_Zmat)) # multiply 1 to make this matrix numeric

ht1 = Heatmap(t(id.mat), 
             cluster_rows = F, 
             cluster_columns = F,
             show_row_dend = F, show_column_dend = F,  # do not show dendrogram
             #show_column_names = F, col = c("white","red"),
             show_column_names = F, col = c("lightgray", "blue"),
             row_names_gp = gpar(fontsize = 6), name="Tested",
             heatmap_legend_param = list(labels = c("Yes", "No"),
                                         title_gp = gpar(fontsize = 6), 
                                         labels_gp = gpar(fontsize = 4)))
draw(ht1)

Run kompute()

We use kompute() to impute the missing Z-scores in the Z-score matrix.

kompute.res <- kompute(BC_Zmat, BC_Pheno_Cor)

Results

The kompute function produces a list consisting of two matrices: i) the Z-score matrix, and ii) the corresponding imputation information matrix. For the measured Z-scores, the information is set to NA, while the imputed Z-scores contain information values ranging between 0 and 1.

head(kompute.res$zmat[,1:3])    # Z-score matrix
#>      BC_BMC/Body weight BC_Body length BC_Bone Area
#> [1,]         -0.6553998     -1.6404679   -0.2415052
#> [2,]          0.1949131     -0.5418602   -0.5630102
#> [3,]         -0.2398292     -0.3960885   -3.0121842
#> [4,]         -0.1703480     -3.6834565   -0.3905398
#> [5,]         -1.0319219      0.6519419   -1.7038746
#> [6,]          1.7459897     -0.5004111    1.2329180
head(kompute.res$infomat[,1:3]) # imputation information matrix
#>           [,1]      [,2]     [,3]
#> [1,]        NA        NA       NA
#> [2,] 0.4360353 0.4494443 0.203262
#> [3,] 1.0000000 0.5220450 1.000000
#> [4,]        NA        NA       NA
#> [5,] 1.0000000 1.0000000 1.000000
#> [6,]        NA        NA       NA

ht2 = Heatmap(t(kompute.res$zmat), 
             cluster_rows = F, 
             cluster_columns = F,
             show_row_dend = F, show_column_dend = F,  # do not show dendrogram
             show_column_names = F, col = c("blue", "red"),
             row_names_gp = gpar(fontsize = 6), name="Z-score",
             heatmap_legend_param = list(title_gp = gpar(fontsize = 6), 
                                         labels_gp = gpar(fontsize = 4)))
draw(ht2)

For more detailed analysis examples demonstrating the effectiveness of the kompute imputation method, please visit our examples page (komputeExamples).

statsleelab/kompute