Xiuwen Zheng

Whether you need Bonferroni correction or not depends on your scientific question. Try `hlaAlleleToVCF()` to convert hlaAllele to PLINK. ``` hlaAlleleToVCF(hla, outfn, DS=TRUE, verbose=TRUE) ```

Update the version of the HIBAG package. CHANGES IN VERSION 1.28.0 ------------------------- o new function `hlaAlleleToVCF()` for converting the imputed HLA classical alleles to a VCF file

No need to use HWE filter in QC, the HIBAG algorithm is robust to the departure of HWE.

Use that probability for a pair of alleles, and a threshold 0.5. Or output all prob. of the full pairs of alleles, ``` hlaPredict(object, snp, type="prob") ```

I have not implemented such function (AA test with continuous outcomes) yet. This feature requirement is added to my to-do list now.

HIBAG.gpu is not formally released yet.

You should check the sample IDs before you run any HIBAG function. See `train.geno_sea730k$sample.id` and `true_b$value$sample.id`.

If you specify a list of SNP IDs in `snpgdsIBDMLE()`, it works.

``` seqVCF2GDS("your_vcf_file.vcf.gz", "your_output.gds") ``` The SNPRelate could read the SeqArray GDS file directly, without converting to SNP GDS.

You misused "genotype.var.name", dosages are always stored in 'annotation/format/DS'. Remove `,genotype.var.name="annotation/format/DS"`