Software

New features and developments in the BLUPF90 software suite

---

Abstract

BLUPF90 is one of the most used software packages in animal breeding and genetics, with over 2500 monthly downloads from worldwide users. This is primarily due to its flexibility and ability to perform various tasks in our field, including variance components estimation (VCE), breeding value estimation, accuracy computation, and genome-wide association, among others. This software suite is dynamic, with weekly updates and new tools being developed annually. Here, we describe all the new features and developments in the BLUPF90 software suite. Recently, we added validationf90, software that performs model validation based on predictive ability and linear regression (LR), with standard errors and confidence intervals, providing a framework for proper validation. We also added rrmebvf90, which is a post-processing tool for random regression and reaction norm models. It computes breeding values, reliabilities, variance components, and genetic parameters for a specific point or a cumulative interval of the longitudinal scale. A third software added, dhglmf90, is in fact a wrapper that allows blupf90+ to estimate genetic heterogeneity in residual variance using the double hierarchical generalized linear model (DHGLM); it jointly estimates genetic effects and (co)variance components for the mean and dispersion components of the model. We recently modified blup90iod3 (now blup90iod3+) to perform VCE with millions of genotyped animals, which is based on Monte Carlo REML (MC-REML) with GEBV simulation under single-step GBLUP (ssGBLUP). The method, named MC-ssGREML, has been successfully tested in various livestock datasets. For breeding value estimation in large genotyped populations, we implemented memory mapping, an efficient method for working with large files by mapping a segment of virtual memory to a portion of a file stored on disk. In this way, the file can be accessed as if it were regular memory, but without occupying big chunks of memory, as large files are not loaded into RAM. We also updated the software to approximate the reliability of GEBV using Tier and Meyer's algorithm and capabilities for random regression and reaction norm models, which allow for approximating genomic reliabilities for point or cumulative GEBV, as well as for an index GEBV. This new software was named accf90GS3. Based on the algorithm to approximate genomic reliability, we developed an approximation to obtain p-values for SNP effects in single-step GWAS for large genotyped populations, which was implemented in postGSf90 and has been tested with over 2.5 million genotyped animals. Recently, we derived and implemented Expectation-Maximization and Newton-Raphson algorithms into blup90iod3 to enable computations of GEBV for multiple categorical and continuous traits under threshold models, a feat previously deemed unattainable. Together, these developments expand the scope and scalability of the BLUPF90 software suite, reinforcing it as a key tool for genomic analyses across species and complex, large datasets.

Keywords: 2026

How to Cite:

Lourenco, D., Bermann, M., Aguilar, I., Legarra, A., Tsuruta, S. & Misztal, I., (2026) “New features and developments in the BLUPF90 software suite”, World Congress on Genetics Applied to Livestock Production Digital Archive 2026(1): 2284867. doi: https://doi.org/10.31274/wcgalp.23619

Rights:

1