Genomic information from 13,777 Brazilian Angus animals born between 1988 and 2023 was provided by the Brazilian Angus Association (Promebo^®, 2020). Animals were genotyped using fourteen different SNP panels with densities ranging from 35 K to 150 K. After combining the SNPs and performing quality control, they were imputed to a density of 78,837 markers using the FImpute software version 3 (Sargolzaei et al., 2014). The reference population for genotype imputation consisted of 835 Brazilian Angus animals born before 2013 and genotyped with high-density SNP panels (150 k SNPs). These animals were selected for their pedigree and phenotypic data to capture the genetic diversity of the population. A rigorous quality control process was applied prior to genotype imputation to ensure data reliability. Samples with low call rates (<90%), extreme heterozygosity, sex mismatches, or duplicates were removed. SNP filtering retained only autosomal markers with CR >98%, MAF >0.03, and Hardy–Weinberg p-values >10⁻⁷, while redundant or highly correlated SNPs were excluded. Details about the imputation process performed in this population can be found in Campos et al. (2022).

In this study, genotypic quality control procedures were implemented in accordance with the analyses performed. For instance, for the identification of ROH, only data with call rate exceeding 95% for both animals and genotypes were used (no animals or SNP markers were excluded from the analysis based on these criteria). For the other inbreeding metrics used in this study, minor allele frequency (MAF <0.05) and Hardy-Weinberg equilibrium (HWE <10⁻⁶) were also used to filter out markers. Ultimately, 78,837 and 71,457 SNP autosomal markers were retained for ROH identification and other analyses, respectively. A principal component analysis referring to the population structure is presented in the Supplementary File 1.

Phenotypic information on birth weight, weaning weight gain, weaning conformation, weaning hair coat, post-weaning weight gain, yearling conformation, yearling hair coat, scrotal circumference, ribeye area, subcutaneous back fat thickness, rump fat thickness, intramuscular fat, and tick count were used for this study. Birth weight was measured right after birth (kg). Weaning weight gain was adjusted to 205 days (kg) and post-weaning weight gain was adjusted to 365 days (kg). Weaning conformation and yearling conformation scores are visual measures of the volume of the carcass, taking into account body length and rib depth. Each animal was assigned a score between 1 and 5, with five being the highest expression of the trait and one being the lowest, relative to its contemporary group. Weaning hair coat and yearling hair coat scores range from 1 to 3, with one indicating a short hair coat, two indicating a medium hair coat, and three indicating a long hair coat. Ribeye area (cm²), subcutaneous back fat thickness (mm), and intramuscular fat (%) were measured using an ultrasound device on the region between the 12th and 13th ribs, transversely over the Longissimus dorsi muscle. Rump fat thickness (mm) was measured by ultrasound on the animal’s rump, between the Gluteus medius and Biceps femoris muscles. Tick counts were performed manually by counting adult female ticks with at least 4.5 mm on one side of each animal (Wharton et al., 1970). One to three subsequent tick counts were obtained between 2012 and 2022. Tick count records were log-transformed to normalize the distribution, and a constant of 1.001 was added to the counts prior to this transformation because log10(1.0) = 0.0 (Cardoso et al., 2015). Scrotal circumference was measured at yearling, using a specific metal tape at the widest point of the scrotum (cm). Contemporary groups were formed by animals from the same farm, sex, year and season of birth, management group, and date of phenotypic evaluation. Data consistency was ensured by checking for outliers in the continuous traits, i.e., data exceeding 3.5 SD above or below the contemporary group average were excluded. Moreover, contemporary groups with fewer than three animals and/or with no variability for the score traits (i.e., weaning conformation, weaning hair coat, yearling conformation, and yearling hair coat) were also removed (Campos et al., 2022). Descriptive statistics for all traits were computed for both the entire population and a subgroup of only genotyped animals (Table 1).

The pedigree information comprises data on 530,327 animals born from 1900 to 2024, with 84,625 classified as founders, 12,069 identified as sires, and 172,602 identified as dams. The longest recorded ancestral path extends up to the 17th generation, with a total of 121,039 animals showing signs of inbreeding. A complete and comprehensive analysis of the pedigree structure is added in the Supplementary File 2.

The PLINK v1.90 software (Purcell et al., 2007) was used to identify ROHs, following the parameters proposed by Mulim et al. (2022). For instance, ROHs were defined using a sliding window of 50 SNPs, a minimum of 30 consecutive SNPs, and a minimum length of 500 kb were required for the region to be identified as ROH. A SNP density of at least 1 SNP per 50 kb was also required, with a maximum gap between consecutive SNPs set at 1,000 kb. A window threshold of 0.05 was applied, allowing for one heterozygous and one missing SNP in the window. After identification, the ROHs were categorized into groups based on length: <2 Mb, 2–4 Mb, 4–8 Mb, 8–16 Mb, and >16 Mb.

Six different inbreeding metrics were used in this study. The first metric was derived from the pedigree data (F_PED). F_PED was estimated using the CFC software (Sargolzaei et al., 2006), employing the methods outlined by Meuwissen and Luo (1992). The equation used for the calculation of this inbreeding metric is defined as: F P E D i = A i i − 1 where A _ii is the diagonal element of the relationship matrix between the animals of the population on the tabular method proposed by Lush (2010). The second metric used in this study is based on the genotype additive variance (F_GRM). Therefore, this approach uses the model proposed by VanRaden (2008) defined as: F G R M = G i i − 1 where G i i = Z Z ′ ∑ j = 1 M 2 p i 1 − p i where G _ii is the genomic relationship matrix, Z is the n x M matrix of centered genotypes for n individuals, and p_i is the reference allele frequency in the population. The third metric was based on the homozygous genotypes observed and expected (F_HOM1), as proposed by Li and Horvitz (1953): F HOM 1 = H exp − H o b s H exp where, H_exp is the expected value for homozygous genotypes and H_obs is the observed value for the homozygous genotypes. Similar to this method, the fourth metric (F_HOM2) used in this study was also based on the homozygous genotypes following the proposed by Yang et al. (2011), which was estimated as: F HOM 2 = 1 − ∑ i x i * 2 − x i ∑ i 2 p i 1 − p i where x_i is the number of reference allele copies of the ith SNP, and p_i is the reference allele frequency in the population.

Given that all previous metrics strongly relied on genotype allele frequency, an additional metric that tries to minimize the effect of allele frequency on the estimation was also tested. This estimation is based on the correlation between uniting gametes (F_UNI). This method follows the model proposed by Yang et al. (2010), which is defined as: F U N I = x i 2 − 1 + 2 p i * x i + 2 p i 2 2 p i 1 − p i where x_i is the number of the reference allele copies of the ith SNP, p_i is the reference allele frequency in the population.

The final metric used in this study was based on the sum of the individual lengths of ROH divided by the total length of the autosomal genome (F_ROH). This approach followed the equation proposed by McQuillan et al. (2008): F R O H = ∑ i = i n f R O H i ∑ j = 1 A h j where f R O H i represents the length of the ROH for the individual ith, n denotes the total homozygous genomic regions of each individual, h(j) is the length of the chromosome jth, and A is the number of autosomal chromosomes (A = 29). Furthermore, for each class of ROH (<2 Mb, 2–4 Mb, 4–8 Mb, 8–16 Mb, >16 Mb), inbreeding estimates were computed by dividing the total sum of the length of ROH segments by the total length of the autosomal cattle genome. The four first genomic inbreeding coefficients were calculated using the PLINK v1.9 software (Purcell et al., 2007). The correlation among the different inbreeding metrics was assessed using the Pearson correlation coefficient (Pearson, 1896), which was estimated using the R software (R Development Core Team, 2009).

Inbreeding depression was estimated, for each inbreeding metric, using a single-trait linear model by regressing the phenotypes on the estimated inbreeding coefficient. The following statistical model was used to analyze birth weight, post-weaning weight gain, weaning conformation, and weaning hair score: y = 1 μ + X b + β F + Z a + W m + C m p e + e where y is the vector of phenotypic records; 1 is a vector of ones; µ is the average of the phenotypes; b is the vector of fixed effects including contemporary group (farm-year-season), animal’s sex group, cow’s age in years, and age of the animals in days (linear and quadratic effects) as covariates; F is the vector of the inbreeding coefficient (F_PED, F_GRM, F_HOM1, F_HOM2, F_UNI, F_ROH, and different groups of ROH classes); and β is the linear regression coefficient for the inbreeding coefficient; a is the vector of random additive genetic effects, with a ∼ (0,A σ a 2 ), where A is the numerator of the pedigree-based relationship matrix, and σ a 2 is the additive genetic variance; m is the vector of random maternal additive genetic effects (not used for weaning conformation and hair score), with m ∼ (0,A σ m 2 ), mpe is the vector of random maternal permanent environment effect with mpe ∼ N (0,I σ m p e 2 ), and e is the random residual, with e ∼ N (0,I σ e 2 ), where σ e 2 is the residual variance. The X, Z, W, and C are the incidence matrices related to b, a, m, and mpe, respectively.

For yearling conformation, yearling hair coat, scrotal circumference, ribeye area, subcutaneous fat thickness, rump fat thickness, and intramuscular fat, the following statistical model was used in the analysis: y = 1 μ + X b + β F + Z a + e where all terms and assumptions were previously defined. For tick count records, which were log-transformed to normalize the distribution, the following model was used: y = 1 μ + X b + β F + Z a + W p e + e where pe is the vector of permanent environmental effects, with pe ∼ N (0,I σ p e 2 ), where σ p e 2 is the permanent environmental variance; and W is the incidence matrix related to pe. All other terms and assumptions were previously defined. Confidence intervals for the β coefficients were calculated as β ± 1.96 × s e β ; and used to recognize which traits had an inbreeding depression statistically different from zero. Variance components estimation was performed using the Average Information Restricted Maximum Likelihood (AIREML) algorithm implemented in the BLUPF90 family programs (Misztal et al., 2014; Lourenco et al., 2020).

The summary of the inbreeding coefficients estimated using the different metrics evaluated in this study is shown in Table 2.

Individual inbreeding coefficients ranged from −0.33 to 0.71, using the different metrics (Table 2). The metrics F_PEDtotal, F_PEDgen, F_GRM, F_HOM1, F_HOM2, F_UNI, and the class of <2 MB of the ROH had the lowest average inbreeding coefficient (0.01). On the other hand, the F_ROH metric showed the highest inbreeding coefficient (0.13 ± 0.03). Figure 1 illustrates the average inbreeding over time based on pedigree data, alongside the number of animals in the pedigree file (Figure 1A). Additionally, it shows the trend of average inbreeding population over the years for the genotyped animals using different inbreeding metrics (Figures 1B,C) and the overlap of the 10% most inbred animals for the genomic inbreeding metrics (1D).

In general, the number of animals in the population increased until 2013, followed by a decline. Conversely, the average of inbreeding, based on pedigree, decreased during this period (Figure 1A). On the other hand, as shown in Figure 1B, the average inbreeding for the genotyped animals increased over the years according to the F_ROH, F_PED, F_HOM1, and F_HOM2 metrics, while the F_GRM and F_UNI metrics showed a decrease. Figure 2 shows the distribution frequency of inbreeding metrics and the correlations among the various inbreeding metrics.

The correlation among different inbreeding metrics ranged from weak (−0.26; between F_GRM and F_2–4Mb) to strong (0.97; between F_ROH and F_HOM2). The metrics that had high correlations among themselves were F_HOM1, F_HOM2, F_UNI, and F_ROH. The F_PED metric showed a weak correlation with all the others, with its highest correlation being observed with the segments of ROH longer than 16 Mb (i.e., 0.41). Negative correlations were observed among F_GRM and the majority of inbreeding coefficients derived from the ROH segments.

Table 3 shows the estimates of inbreeding depression for all traits analyzed in this study, using the different inbreeding metrics. Interestingly, the extent of inbreeding depression varied across the traits evaluated (Table 3). While significant effects were evident for several key growth and carcass component traits, other traits such as tick count did not exhibit statistically significant impact from inbreeding depression.

The estimation of inbreeding depression on the traits analyzed in this study showed a predominantly unfavorable effect, with increased inbreeding having the most significant unfavorable impact on rump fat thickness, where a 1% increase in the inbreeding coefficient decreases the performance of the trait up to 1.13%, considering the F_8–16Mb metric. Conversely, favorable associations with increasing inbreeding were observed for intramuscular fat (F_HOM2) and birth weight (F_2Mb). In summary, the inbreeding depression on different traits ranged from −1.13% in the F_8–16Mb metric for rump fat thickness to 0.83% in the F_2Mb metric for birth weight. The b values and their respective standard deviations can be found in Supplementary File 3.

The accurate quantification of inbreeding is fundamental for developing effective breeding designs strategies, as it directly impacts the interpretation and management of inbreeding depression. While traditional pedigree-based inbreeding coefficients (FPED) provide estimates based on recorded ancestry, they can be compromised by errors or incompleteness in pedigree data, potentially overlooking cryptic relatedness, particularly in populations with shallow or poorly documented relationships. In comparison, genomic-based metrics, such as FROH, offer a more precise alternative by directly estimating realized autozygosity from an individual’s genome. This approach provides higher resolution and greater power to detect recent inbreeding events or segmental autozygosity. Among genomic metrics, we believe FROH is particularly valuable due to its ability to distinguish between recent and ancient inbreeding. For instance, our results suggest that longer ROH segments, indicative of recent inbreeding, seems to be associated with more severe inbreeding depression. Thus, we believe that for producers and breeding programs, focusing on minimizing the accumulation of long ROH segments might be a pragmatic initial approach, while still monitoring overall FROH from various length classes to manage long-term genetic diversity.

Despite the valuable insights this study offers regarding inbreeding effects on key traits in Brazilian Angus cattle, several limitations warrant consideration. The accuracy of genomic inbreeding coefficients is highly dependent on factors such as SNP panel density, marker distribution, and overall genotyping quality. In this context, we acknowledge that genotype imputation, while a valuable tool, can introduce certain limitations and potential biases in our study. For instance, the underestimation of rare variants or the potential for imputation errors to slightly affect downstream analyses like ROH calling, particularly for very short segments. We emphasize that while we employed best practices to minimize such issues, the inherent limitations of imputation should be considered when interpreting the results of this paper. Additionally, the availability of sparse phenotypic data for certain traits, for example, tick count, reduced the statistical power for detecting inbreeding depression. It is also important to note that the analytical models used in this study did not explicitly account for the individual effects of deleterious recessive alleles or complex allele interactions, such as dominance or over dominance, which can likely modulate the expression of inbreeding depression.

The integration of genomic inbreeding metrics into breeding programs offers a powerful complement or alternative to traditional pedigree-based approaches, particularly in populations like Brazilian Angus, where pedigree information may be incomplete. These metrics enhance the efficiency of advanced strategies such as optimum contribution selection and genomic-based mating allocation, helping breeders balance genetic gain with the control of inbreeding. Moreover, monitoring ROH can identify individuals at risk of inbreeding depression and guide more informed mating decisions, promoting long-term genetic diversity and adaptability under tropical conditions. While our study focused more on carcass and growth traits due to data availability, we acknowledge the critical importance of reproductive traits—highly sensitive to inbreeding depression—and recommend that future research prioritize their inclusion to fully assess the impact of inbreeding and refine genetic management strategies.