Imputor is a very simple library for imputing missing variants in a genotype dataset (i.e. 23andMe)

Users can run Imputor as either a standalone command line tool or import it into their python library

Requirements ***********

The imputation uses an LD approach and in order to calculate the LD a reference dataset is required (usually the 1000 Genomes HapMap dataset). This HapMap dataset has to be provided as an HDF5 file and should only contain bi-allelic SNPs. A version can be downloaded here.

There are 6 steps to finish in the imputation pipeline. Some of the steps only have to be done once or once for each genotyping plattform and dataset respectively Each step maps to a subcommand of the command line script. To get information about the various subcommands the user can run:

imputor -h

Currently risk prediction only works reliably with European indviduals. So we first need to remove all non-european individuals from the 1000 genomes dataset and also remove the monomorphic SNPs. This can be done with the prepare command. This only has to be done once

imputor prepare 1000_genomes.hdf5 1001_genomes_unrelated.hdf5

The genotypes from 23andMe and others are usually provided in text or csv format. We first need to convert them into an HDF5 format for the downstream analysis:

imputor parse genotype.csv genoptype.hdf5

The genotypes have to be converted from the NT form to a binary form. To speed up this conversion a nucleotide map is constructed. The map is constructed from the HapMap dataset that was created in Step 1. This has to be done once for each genotype version and platform respectively. For each version/platform the steps from now on have a separate pipeline!

imputor nt_map 1000_genomes_unrelated.hdf5 genoptype.hdf5 nt_map.pickled

For the imputation LD files have to be created. This has to be done once per genotype version and platform respectively. The user can specify an optional window_size for the LD calculation

imputor calc_ld 1000_genomes_unrelated.hdf5 nt_map.pickled FOLDER_FOR_LD_FILES

In the last step before imputation the genotype file has to be converted from NT form to binary form. This is done using the the nucleotide encoding map that was generated in Step 3:

$ imputor convert genotype.hdf5 genotype_encoded.hdf5 nt_map.pickled

Using the LD files the missing SNPs can be imputed for an existing genotype:

$ imputor impute genotype_encoded.hdf5 FOLDER_FOR_LD_FILES

Of course, the recommended installation method is pip:

$ pip install imputor

Thanks for checking this library out! We hope you find it useful.

Of course, there's always room for improvement. Feel free to open an issue so we can make Imputor better.