README.md

In GaoGN517/689_SNPFastImpute_Mac: Perform fast missing value imputation for single nucleotide polymorphism data.

SNPFastImputeMac

The goal of SNPFastImputeMac is to impute missing values in SNP data files.

Installation

You can install the development version of SNPFastImpute from github using devtools:

devtools::install_github("GaoGN517/689_SNPFastImpute_Mac")

This is a duplicated version of my previously developed package on windows system. I met a problem when install it to mac system. So I copied all necessary documents and folders from the previous package to this new one. This one has no such an error when install on mac system.

Basic Example

This is a basic example which shows you how to solve a common problem:

library(SNPFastImputeMac)

## Read a vcf file as a matrix
## filename <- "data/Test.vcf" ## your file path
## vcf_df <- read.table(filename) ## read in your vcf data 
## Here we just load the dataset in vcf format.
data(vcf_df)
output_df <- vcf2df(vcf_df)
## There would be warning message for NAs, which is caused by adding missing positions
## of SNPs, which is what we should do. So here NAs does not mean problem. 

## Introduce some missing values into the original matrix to test the performance
data("SNP_orig_sub")
## this is the SNP matrix with the original SNP types.
## original NA ratio is 0.018

## Make the final missing ratio to be 20%
SNP_NA_df02 <- NA_Generator(SNP_orig_sub, 0.2)
## This is an object list of four elements, 
## SNP_NA_df, 
## NA_percent_orig,
## NA_percent_generate,
## NP_generate_positions

## Make another object with final missing ratio to be 5%
SNP_NA_df005 <- NA_Generator(SNP_orig_sub, 0.05)

## Now we have the original matrix, the two objects with additional missing values.
ls()
# [1] "SNP_NA_df005" "SNP_NA_df02"  "SNP_orig_sub"

## We can use the following function to create an object that can be used in the imputation 
## function for each SNP in the matrix.
Create_Single_SNP_Object(SNP_NA_df005$SNP_NA_df, 2, size = 20)

## The main function of this package is to use the imputation function to fill in the missing
## values.

## We can just do the filling for the original matrix
system.time(
  predict_df <- Impute_GenoType_XGBoost(SNP_orig_sub, size = 10)
)
##    user  system elapsed 
##   1.898   0.027   1.949 

## We can also perform the filling on the matrix which we introduced additional 
## missing values. And then see how our method performed on predictions. 
system.time(
  df_fill02 <- Impute_GenoType_XGBoost(SNP_NA_df02$SNP_NA_df)
)
##   user  system elapsed 
##   10.202   0.090  10.455 
system.time(
  df_fill005 <- Impute_GenoType_XGBoost(SNP_NA_df005$SNP_NA_df)
)
##    user  system elapsed 
##   8.900   0.156   9.445 

NA_positions02 <- SNP_NA_df02$NP_generate_positions
NA_positions005 <- SNP_NA_df005$NP_generate_positions

classification_error(SNP_orig_sub, df_fill02, NA_positions02) ## 0.076
classification_error(SNP_orig_sub, df_fill005, NA_positions005) ## 0.049

GaoGN517/689_SNPFastImpute_Mac documentation built on Dec. 8, 2019, 12:33 a.m.