Roadmap

[Hua-Zhou]

This issue documents the implementation roadmap for iterative solvers.

• Convert GT data to Plink format (SnpData)
• Subset VCF records according chromosome and position range
• Subset VCF records according to marker index or IDs
• Subset VCF samples according to individual index or names
• Convert methods for VCF.Reader, e.g., convert(Matrix{T}, reader::VCF.Reader) will convert all records from the current position of reader to a matrix of type Matrix{T}
• Copy methods for VCF.Reader, e.g., copy(A::AbstractMatrix{T}, reader::VCF.Reader) will fill the columns of A the GT data from the current position of reader

[biona001]

Below is a (tentative) list of features I need for MendelImpute:

• filter function based on sample and record index
• convert_ht function to import VCF files into a numeric matrix where columns are haplotypes
• convert_ds function to read dosage into a numeric matrix
• convert_vcf function to convert phased genotype matrix back to a VCF file. This is not implemented explicitly but is support with general write methods. See MendelImpute's impute.jl

I will try to implement them in the next 1~2 weeks.. Are there any caveats I need to be aware of? In particular, I'm not sure what is the best way to filter based on sample index due to the data structure of VCF.Record.

[biona001]

Here are a few more desired routines typically needed for quality control:

• splitting multi-allelic calls into different records
• left-aligning indels
• removing 'junk' variants that have a high statistical probability of being false-positives via the QUAL score, which is negative log 10 Phred-scaled quality.

They are mentioned here

[kose-y]

see https://github.com/vcftools/vcftools/blob/master/src/cpp/variant_file_format_convert.cpp for vcf to plink conversion.

[kose-y]

VCF to PLINK is implemented in SnpArrays.jl.

[biona001]

In the next few days, I will add:

• Calculation of Hardy Wienburg equilibrium p-value using Fisher's exact test (reference)
• GRM calculation, with emphasis on treating missing data (reference)

[Hua-Zhou]

👍