Marker assisted breeding (MAB) is a crop and animal breeding method that selects individuals by identifying molecular markers associated with important traits. Genome-wide association studies (GWAS) are a MAB method that uses linkage disequilibrium (LD) to associate traits of interest with single nucleotide polymorphisms (SNPs) which can function as molecular markers (Klein et al. 2005). GWAS is an effective tool for improving simple monogenic traits, where a few genes have a large effect. These simple traits can be applied through gene pyramiding which stacks the molecular markers identified by GWAS into a single genotype (Servin et al. 2004). Improving complex polygenic traits, where there are many genes with a small or varied effect on the trait of interest, is more effective using genomic selection (GS). GS is a MAB method that predicts trait expression in the form of genomic estimated breeding values (GEBVs). These GEBVs can then be used to identify optimal individuals for breeding (Meuwissen et al. 2001).

Local haplotypes are multiallelic molecular markers consisting of a group of SNPs that are inherited together due to high levels of linkage disequilibrium (Belzile et al. 2020). The use of haplotypes can improve the results of GWAS and GS analyses by accounting for local epistatic effects. Haplotype-based GWAS has greater power to detect QTLs, lower false discovery rate, can map QTLs to a smaller region, and identify a greater number of novel QTLs than conventional SNP-based GWAS (Pryce et al. 2010, Chen et al. 2021, Helal et al. 2021, Wu et al. 2022). Haplotypes can improve trait prediction accuracy in GS, especially for traits with high haplotype epistatic heritability (Sallam et al. 2020, Bian et al. 2021, Weber et al. 2023).

There are two main categories of local haplotyping tools, gene-centric and clustering tools. Gene-centric tools define a haplotype as a small region centered around a gene and can use a fixed length to determine the haplotype region around the gene, e.g. Candihap (Li et al. 2023) or RFGB (Wang et al. 2020), or they can define the haplotype based on LD in the region surrounding the gene, e.g. HaploMiner (Tardivel et al. 2019). Clustering local haplotyping tools such as crosshap (Marsh et al. 2023) arranges molecular markers into groups based on LD. Crosshap defines local haplotypes in a small region of the genome and visualizes the results. These gene-centric and clustering local haplotyping tools are designed to investigate a small region in greater detail. They are best used once a candidate gene, or a region of interest has been identified and are not suitable for defining genome-wide haplotypes for use in haplotype-based MAB. In contrast, HapFM clusters SNPs into local haplotypes and then performs GWAS (Wu et al. 2022). Wu et al. (2022) found consistently higher mapping power using HapFM when compared to SNP-based GWAS algorithms, demonstrating how haplotypes can improve MAB results. However, HapFM is limited to GWAS analyses and cannot be used for other MAB approaches.

Here we present HaploVar, a local haplotyping tool that clusters SNPs into LD-based haplotypes, identifies haplotype variants and formats the output to be compatible with common GWAS and GS tools. The purpose of HaploVar is to improve GWAS and GS pipelines by supporting haplotype-based MAB methods.

HaploVar accepts genotype data in VCF format and a pairwise LD matrix and uses these to define local haplotypes, identify the variants for each haplotype, and to output these local haplotype variants in one of six formats for input into a wide range of GWAS and GS tools. The LD matrix can be calculated using tools such as PLINK (Purcell et al. 2007, https://zzz.bwh.harvard.edu/plink/ld.shtml) or the ld function of snpStats (Clayton 2025). HaploVar has two sets of functions: local_haplotypes, which defines local haplotypes and produces a list of local haplotypes and their SNPs; and haplotype_variants, which identifies haplotype variants, and formats the output for GWAS or GS analyses. These function sets are discussed in more detail in sections 2.1 and 2.2, respectively. In each function set there is the base function, a function with the suffix _globally which runs the base function over a list of VCF and corresponding LD matrices, and a function with prefix collate_ which collates output from the base function. These functions are designed to optimize use of computational resources. HaploVar can be installed using the following R code:

install.packages(“HaploVar”)

library(HaploVar)

HaploVar is designed to improve MAB pipelines by the preparation of haplotype-based data for GWAS and GS. HaploVar takes SNP and LD data, calculates local haplotypes, identifies haplotype variants, and formats the output for use in GWAS and GS tools. An example application of HaploVar can be found in Note 2, available as supplementary data at Bioinformatics online. Comparison of HaploVar with other haplotyping tools is shown in Note 3, available as supplementary data at Bioinformatics online. Note 4, available as supplementary data at Bioinformatics online shows a parameter sensitivity analysis. Using haplotypes in MAB studies can improve power and trait prediction accuracy.

Current local haplotyping tools focus on the downstream analysis and fine-mapping of GWAS results or are limited to only haplotype-based GWAS. HaploVar is the first tool designed utilize the benefits of local haplotypes for both haplotype-based GWAS and haplotype-based GS.