The experimental data consisted of 3,605 Brangus cattle, 20,359 Angus cattle, and 483 Brahman cattle. The latter two ancestral breeds were used as the reference populations for estimating GBC for Brangus cattle. All the animals were genotyped with a GeneSeek Genomic Profiling (GGP) bovine 50 K V1 (version 1) chip, except 349 Brahman cattle were genotyped with an Illumina 777 K bovine SNP chip (Table 1). The genotypes were extracted from the Neogen GeneSeek genotyping databases representing samples shared between the Neogen global laboratories. The SNP map positions were based on the UMD 3.1 reference bovine genome assembly (Merchant et al., 2014).

Reference SNPs were selected from the common set between the GGP bovine 50 K V1 chip and the Illumina 777 K bovine SNP Beadchip. The data cleaning removed SNPs with a call rate of less than 95%, SNPs on the two sex chromosomes, and SNPs without map position. SNPs violating the Hardy-Weinberg equilibrium (p < 1.0E-8) were also excluded. For SNPs with greater than 0.99 correlations on each chromosome, only the one with higher or the highest minor allelic frequency was kept. When there were ties in allelic frequencies, a random one was taken. The final SNP set retained 41,672 common SNPs for the subsequent analyses. In each ancestral population, outlier individuals were excluded as those with (−2)log (likelihood) exceeding a given cutoff value (i.e., 2.0 by default) (He et al., 2018). This test excluded 37 Angus and 26 Brahman animals from the reference populations. The means and standard deviations of allele A frequencies of the SNPs after data cleaning in the three populations are shown in Table 1. On average, Angus cattle had a higher allele A frequency (0.492) than Brahman cattle (0.431–0.439). The average allele A frequency for Brangus cattle was 0.477, which fell between the two ancestral populations yet closer to Angus. The Angus population had a smaller standard deviation of allele A frequency than the Brahman population. The Brangus cattle had a smaller standard deviation of allele A frequencies than the two ancestral populations (Table 1).

Breed compositions were estimated for individual Brangus animals using the admixture model by Bansal and Libiger (2015) (BL-Admixture). This method utilizes the same form of likelihood model as in the STRUCTURE (Pritchard et al., 2000) and ADMIXTURE (Alexander et al., 2009) software packages. However, the BL-Admixture model runs faster. In particular, STRUCTURE takes a Bayesian approach and relies on a Markov Chain Monte Carlo (MCMC) algorithm to sample the posterior distribution, which can extremely computationally intensive with large data. STRUCTURE and Admixture represent unsupervised analysis of the ancestry of multiple individuals and jointly estimate allele frequencies for the ancestral populations and the relative contribution of each ancestral population to each individual’s genome. The BL-Admixture estimates the ancestry for a single individual using information about allele frequencies at a large number of loci for multiple reference populations. The latter allele frequencies are obtained from previous unsupervised admixture analysis, or simply from the reference population assuming no genetic drift and selection after these breeds were formed. We took the latter approach.

Consider one biallelic locus (say j) genotyped on an animal (say i). Assume that allele frequencies on this locus are known for each ancestry breed. The admixture model postulates that a progeny’s genotype at this locus is determined according to an allelic frequency as a weighted average of T reference (ancestry) breeds: f i j = ∑ t = 1 T w i t q t j , where q t j is the frequency of allele B for the jth SNP in the tth reference population,; w i t is the corresponding weight. Then, under the assumption of Hardy-Weinberg equilibrium, the probabilities of observing each genotype ( g i j ) on this animal are the following: Pr g i j | f i j = 1 − f i j 2   g i j = 0 2 f i j 1 − f i j   g i j = 1 f i j 2   g i j = 2 (1)

Now, consider j = 1 , … , M SNPs genotyped on this animal, assuming their mutual independence. The log-likelihood computed for all the M SNPs measured on the ith animal, denoted by l i , is as follows. l i = ∑ j = 1 M ⁡ ln Pr g i j | f i j = ∑ j = 1 M g i j ⁡ ln ⁡ ⁡ f i j + 2 − g i j ln ⁡ ⁡ 1 − f i j + C (2) where C = ∑ j = 1 M ln 2 g i j . Note that the assumption of mutual independence between SNPs does not hold precisely due to linkage or/and random associations between them. Nevertheless, the admixture model is often robust to this assumption violation because the percentage of SNPs in high LD are rare when using moderate to high-density SNP panels (He et al., 2018; Li et al., 2020).

The admixture coefficients, W i = w i 1 … w i T ′ , taken to the GBC of animal i attributable to the reference (ancestral) breeds, are obtained using the Broyden-Fletcher-Goldfarb-Shanno (BFGS) method under the restrictions that w i k ≥ 0 ; ∑ k = 1 T w i T = 1 (Nocedal and Wright, 2006). BFGS is a powerful, Quasi-Newton second derivative line search family method to solve non-linear optimization problems. Optimizing the likelihood function of BFGS iteratively removed the non-zero mixing coefficient, which did not significantly improve the model fitting, thus obtaining a concise set of individual mixture coefficients. In the admixture model, the value of each admixture coefficient is bounded between 0 and 1, and the sum of admixture coefficients (GBC) computed for each animal is one under the assumption of 100% genetic contributions by the T ancestral breeds to each animal.

Breed compositions for the Brangus animals were estimated genome-wide, per chromosome, and per chromosomal segment, respectively. The number of reference SNPs per chromosome varied from 712 (chromosome 25) to 2,568 (chromosome 1), and the average distance between SNPs ranged from 0.05 to 0.07 Mb. SBC were obtained on three window sizes: 1 Mb, 5 Mb, and 10 Mb, respectively. The widow sizes were taken arbitrarily yet still based on two factors, the average length of gene in the bovine genome and the minimum number of SNPs to give stable estimates. To minimize the errors in the estimated SBC due to insufficient SNP coverage, chromosomal segments with less than five SNPs were excluded from computing SBC. There were 2,522 1-Mb segments, 518 5-Mb segments, and 269 10-Mb segments, respectively. The average number of SNPs per segment ranged from 15.2 to 18.8 (1 Mb), 72.0 to 91.1 (5 Mb), and 132.0 to 169.1 (10 Mb), respectively (Table 2). Overall, the average R ² LD per segment decreased as the window size increased, ranging from 0.107 to 0.164 on 1 Mb chromosomal regions, from 0.062 to 0.104 on 5-Mb chromosomal regions, and from 0.050 to 0.081 on 10 Mb chromosomal regions (Table 2). The R ² LD on 1-Mb regions also had a larger average standard deviation (0.05) than those on 5-Mb and 10-Mb regions (0.02).

K-means clustering (Jain, 2010) was conducted on the GBC for the 3,605 Brangus cattle to reveal possible population stratifications, where K = 2 and 3, respectively. Initially, the number of clusters K was specified a priori, and randomly assigned all the animals to each of K distinct, non-overlapping clusters as their initial clusters. Then, the K-means algorithm computed the cluster centroid for each cluster, which is a vector of genotype means for the M reference SNPs, and it re-assigned each animal to the cluster whose centroid was the closest. The last two steps proceeded iteratively till the total within-clustering variation, defined by the sum of all the pairwise squared Euclidean distance in each cluster and summed over all the K clusters, was minimized as much as possible (Hartigan and Wong, 1979): m i n i m i z e ∑ k = 1 K 1 n k ∑ i , i ′ ∈ C k ∑ j = 1 M g i j − g i ′ j 2 where C k stands of cluster k, and n k is the number of animals in the kth cluster. The K-mean clustering analysis was implemented by the “stats” R package.

Gene ontology (GO) term enrichment analysis, or gene set enrichment analysis, was conducted using the “gprofiler” R package with genes identified in 1-Mb chromosomal regions featuring either ancestral breed. This R package performed functional gene enrichment analysis on the input gene lists, mapped genes to known functional resources, and detected statistically significantly enriched terms. Briefly, chromosomal segments satisfying Angus SBC ≥0.75 or Brahman SBC ≥0.50 were extracted. Both cutoff thresholds represented equal upward GBC deviations (i.e., 12.5%) from their expected values. Then, the gene information was extracted by querying in the Ensembl database according to their chromosomal locations on the selected chromosomal segments. Finally, GO clustering analysis was performed on the gene list generated with the different filters. The annotation databases included Gene Ontology–biological processes, cellular components, molecular function (http://geneontollogy.org/) (Ashburner et al., 2000), and Kyoto Encyclopedia of Genes and Genomes (KEGG; http://www.genome. jp/kegg/) (Kanehisa et al., 2016). Other relevant R packages used in this study included “org. Bt.e.g., db”, “biomaRt” (Smedley et al., 2015), and “clusterProfiler” (Yu et al., 2012). Gene lists were extracted using the “getBM” function in the “biomaRt” R package. Briefly, given a set of filters (e.g., the chromosome number, the start and end positions), it retrieved attributes of all genes in this interval from the BioMart database. The attributes information includes “ensembl gene id”, “chromosome number”, “gene start position”, “gene end position”, and “gene description”, etc. Finally, gene extraction was performed on all eligible SBC fragments, and all extracted genes were merged to form a gene list after removing duplication. Enrichment analysis was then performed. All the QTLs were queried and aggregated in the QTLdb database (Release 49, https://www.animalgenome.org/cgi-bin/QTLdb/BT/index) (Hu et al., 2022).

The estimated GBC for the 3,605 Brangus cattle, on average, were 70.2% Angus and 29.8% Brahman (Table 3), which significantly deviated from the officially expected values (i.e., 62.5% Angus and 37.5% Brahman) (p < 2.2e-16). The 95% confidence intervals were 61.1%–85.8% Angus and 14.2%–38.9% Brahman. There were 459 (12.7%) Brangus cattle with Angus GBC ≥80.0% and 19 (0.5%) Brangus cattle with Angus GBC ≥90.0%. Similarly, elevated Angus GBC for the Brangus animals were documented in some previous studies. For example, Paim T. D. P. et al. (2020) estimated that Brangus were 70.4% Angus and 29.6% Brahman based on high-density SNP genotypes (777,962 SNP, BovineHD Beadchip, Illumina, San Diego, CA, United States). Li et al. (2020) showed that Brangus cattle was 69.8%–70.5% Angus and 29.5%–30.2% Brahman based on multiple models, including an admixture model, linear regression, and ridge-regression BLUP, each with a selectively uniform 20 K SNP panel. Wang et al. (2020) proposed using regularized admixture models to estimate GBC for purebred animals to deal with the so-called “Impure Purebred Paradox”, a phenomenon suggesting a higher-than-expected false-negative rate in the identification of purebred animals. They showed that Brangus were, on average, 71.1%–77.1% Angus and 22.9%–28.9% Braham based on various regularized admixture models. Using path analysis model, Wu et al. (2020) showed that Brangus cattle were 68.2%–71.8% Angus and 28.2%–31.8% Brahman. Hence, regardless of the population sizes and the statistical methods used, these studies have consistently pinpointed that significantly higher-than-expected Angus breed proportions in Brangus cattle.

Multiple reasons are likely responsible for the elevated Angus lineage in Brangus. Firstly, population stratification could exist with Brangus, given the fact that the IBBA approved the creation of UB and UR animals in 2005. The density plots of the estimated Angus or Brahman breed proportions in the 3,605 animals were bimodal, which served as preliminary evidence for the Brangus population stratification (Figure 1). Then, K-means clustering was conducted to partition the 3,605 animals into two and three clusters, respectively (Table 3). With K = 2 (i.e., two clusters), the B-1 cluster with a higher average Angus GBC (81.6%) consisted of 713 (19.8%) Brangus cattle, whereas the B-2 cluster with a lower average Angus GBC (67.4%) included 2,892 (80.2%) animals. With K = 3 (i.e., three clusters), the cluster (C-1) with the highest average Angus GBC (82.1%) had 640 (17.8%) Brangus animals. The cluster with the lowest average Angus GBC (67.2%) included 2,584 (71.7%) Brangus animals. In both sets of clustering results (K = 2 versus K = 3), the average Angus GBC for animals in the top Angus composition clusters (B-1 versus C-1) agreed approximately with each other (81.6% versus 82.1%), and they corresponded roughly to the expected Angus breed proportion (81.25%) for the ½ UB animals. We thus suspected that these animals could be the ½ UB animals. Using a path analysis approach, Wu et al. (2020) confirmed that the ½ UB animals were, on average, 81.25% Angus and 18.75% Brahman. The path analysis decomposed the relationships between the ancestors and the composite animals into direct and indirect path effects. The above percentage only accounted for direct effects by the path-analysis interpretation, assuming a zero correlation between the two ancestral breeds (Wu et al., 2020). In reality, however, Angus and Brahman cattle are connected due to sharing common remote ancestors, though the correlation can be low. For example, the correlation of allele A frequencies ranged from 0.05 to 0.10 between the two ancestral breeds, subject to the SNP panel sizes. If considering the indirect path effects, the actual Angus breed proportions for the ½ UB animals could be slightly higher. On the other hand, the two majority clusters, B-2 and C-3, had roughly comparable averages of Angus GBC (67.4% versus 67.2%). These were likely Brangus cattle without the mixture of ½ UB animals. The average Angus breed proportion for non-UB Brangus was 67.4% based on the clustering analysis with K = 2. The 95% confidence interval of Angus breed proportions in Brangus was between 60.4% and 73.5%. Hence, there was, on average, an approximately 5% deviation of GBC toward Angus breed proportions in non-UB Brangus cattle, possibly resulting from selecting Brangus cattle for phenotypes where Angus has advantages. Without selection, genomic-estimated breed compositions would agree approximately with the expected ratios (e.g., Funkhouser et al., 2017; Gobena et al., 2018). Note that ½ UB animals are not precisely registered Brangus. They are officially given a “UB” prefix for registration purposes and are shown on their registered IDs. It is also worth mentioning that the B-2 (or C-3) cluster could include some advanced UB crosses because they had comparable Angus breed proportions as those in the non-UB Brangus cattle in clusters B-2 and C-3. In early 2013, the IBBA further approved the breeding-up of UB (or UR) cattle to registered Brangus. The rule states that an IBBA-registered Brangus sire or dam mated to an IBBA-registered UB or UR sire or dam that results in at least 7/8th (87.5%) Brangus also qualifies as a registered Brangus. For example, mating a ¾ UB animal (i.e., progenies derived from crossing ½ UB animals with Brangus) to a registered Brangus resulted in 7/8 UB animals, which meets the 87.5% Brangus makeup. On average, a 7/8 UB animal was 67.2% Angus. Finally, the average Angus GBC for the animals in cluster C2 was 70.9%, which roughly corresponded to the expected Angus lineage (71.9%) for ¾ UB animals.

The ancestral breed proportions were estimated by chromosomes in the 3,605 Brangus (Table 4). Overall, the average estimated CBC per chromosome varied substantially, from 56.9% (chromosome 5) to 79.8% (chromosome 15). Converse to the Angus CBC, the average Brahman CBC was the lowest on chromosome 15 (20.2%) and the highest on chromosome 5 (CBC 43.1%). We noted that the estimated ancestral breed proportions were heavily distributed toward Angus, which agreed with the breeding target for greater Angus “blood” than Brahman. There were 27 chromosomes with Angus CBC greater than 62.5%, but only two chromosomes (5 and 20) had Brahman CBC greater than 37.5%. Our results coincided with a previous study by Paim T. D. P. et al. (2020). They showed that chromosome 15 had the highest Angus proportion (84.7%), and chromosome five had the largest Brahman proportion (43.7%). Still, there were some differences. Paim T. D. P. et al. (2020) used principal component analysis to describe the population relationships. They showed that the first principal components (PC1), which accounted for ancestral breeds, were uniformly distributed between the two ancestral breeds when evaluated on chromosomes 16, 25, and 29. Because the average Brahman breed proportion is expected to be 5/8 (not 5/5), a uniform distribution of PC1 for Brangus between the two ancestral breeds would suggest equal ancestral genomic proportions. Hence, their results were an indication of significant deviates in breed compositions of Brangus toward Brahman on these three chromosomes. A principal analysis is a popular feature-reducing technique for analyzing large, high-dimension data, yet ignoring the detailed information of individual features, which were local ancestral contributions on specific chromosomal regions. In the present study, we directly evaluated breed compositions by chromosomes. Our results showed that only chromosomes 5 and 20 had less than 62.5% Angus breed proportions on average, which agreed with Paim T. D. P. et al. (2020). But all the other chromosomes (including 16, 25, and 29) had an average Angus CBC greater than 62.5%, meaning they deviated toward Angus instead. These differences could also arise from sampling biases, because the two ancestral breeds used by Paim T. D. P. et al. (2020) were small. There was a high chance that the sampled allelic frequencies may deviate substantially from the actual allelic frequencies. Furthermore, because the variance of an allelic frequency is inversely proportional to the population size (Falconer and Mackay, 1996), the estimated allelic frequencies for ancestral breeds were also subject to large deviations in small samples.

Despite the large chromosome-by-chromosome variability for ancestral breed proportions, GBC estimated by the mean of the average CBC across the 29 chromosomes per animal (denoted by C_GBC) agreed roughly with the average GBC in the 3,605 Brangus cattle. The mean (standard deviation) of Angus C_GBC for the 3,605 Brangus cattle was 70.9% (6.7%). The mean (standard deviation) of Brahman C-GBC was 29.1% (6.7%). The correlation between C_GBC and GBC for the 3,605 Brangus animals was 0.99 (See Supplementary Figure S1A). Within each subpopulation, the average CBC approximately agreed to (or slightly larger than) the corresponding GBC on average. For example, in the two clusters obtained by the K-means clustering with K = 2, the cluster for the mixed UB animals had, on average, 82.5% Angus and 17.5% Brahman breed compositions, and the cluster for non-UB animals had, on average, 68.2% Angus and 31.8% Brahman breed compositions.

Ancestral breed compositions were evaluated on 1-Mb, 5-Mb, and 10-Mb windows on each chromosome (Table 5). Like CBC, the mean and range of the estimated SBC per segment varied substantially with chromosomes (Figure 2). Regardless of the window sizes, all the chromosomes, except 5 and 20, had the mean Angus SBC per chromosome exceeding 62.5%, and only chromosomes 5 and 20 had greater than 37.5% Brahman SBC. The average Angus SBC was the largest on chromosome 15 (76.9%–81.4%) and the smallest on chromosome 5 (57.6%–59.7%) (Table 5). Complementary to the Angus SBC, the Brahman SBC was the largest on chromosome 5 (40.3%–42.4%) and the smallest on chromosome 15 (18.6%–23.1%) (Table 5). Nevertheless, the segment with the maximum Angus proportion appeared on chromosome 18 (94.0%–95.9%), although chromosome 15 had the largest average Angus proportion. The maximum Brahman proportion segment was found on chromosome 2 (74.4%) when evaluated on 1 M windows and chromosome 5 (62.2%–64.4%) on 5-Mb and 10-Mb windows. The overall average of Angus SBC across the 29 chromosomes in the 3,605 Brangus cattle were 69.2% (1-Mb windows), 72.3% (5-Mb windows), and 72.0% (10-Mb windows), respectively. The overall average of Brahman SBC across the 29 chromosomes was 30.8% (1-Mb windows), 27.7% (5-Mb windows), and 28.0% (10-Mb windows), respectively. These overall averages of genomic breed compositions by averaging Angus and Brahman SBC across all 29 chromosomes per animal (denoted by S_GBC) roughly agreed with the average Angus and Brahman CBC (70.9% Angus versus 29.1% Brahman; Table 4) and the average GBC (70.2% Angus versus 29.8% Brahman; Table 3). The correlation between Angus S_GBC and GBC for the 3,605 Brangus animals was greater than 0.99, regardless of the window sizes (See Supplementary Figures S1B–D). Overall, the variation was larger when evaluated on a smaller interval than a larger interval (Figure 3). The Angus SBC on chromosome 15 varied between 68.6% and 89.7% when assessed with a 5-Mb window size (Table 5). The range became smaller (70.5%–86.6%) when evaluated with a larger window size (i.e., 10-Mb), and the range became larger (48.6%–91.1%) when assessed with a smaller window size (e.g., 1-Mb). Similar trends generally held on all the chromosomes (Table 5). The Brahman SBC showed precisely opposite trends.

Ancestral breed proportions per segment varied considerably on each chromosome (Supplementary Figure S2). For example, the Brahman SBC on chromosome one were 4.4%–60.2% (1-Mb windows), 4.6%–47.8% (5-Mb windows), and 7.4%–46.6% (10-Mb windows), respectively (Table 5). There were segments with high Angus proportions and high Brahman proportions, respectively, on the chromosomes. For example, the average Angus SBC in 1–5 Mb and 6–10 Mb segments on chromosome 15 were 85.5% and 89.7%, respectively, which were substantially higher than the CBC for that chromosome. The average Angus SBC computed in the 40–75 Mb segment was between 68.6% and 79.1%, still higher than the officially expected Angus proportion (62.5%). Approximately 95% of the 2,522 1-Mb chromosomal segments had, on average, between 45.0% and 90.0% Angus breed proportions and between 10.0% and 65.0% Brahman breed proportions (Table 6). Likewise, approximately 90% of the total 5-Mb (or 10-Mb) chromosomal segments had between 50.0% and 90.0% Angus breed proportions and between 10.0% and 65.0% Brahman breed proportions (Table 6).

The variability of ancestral breed compositions per chromosome or segment could be a direct effect of selection. For example, Goszczynski et al. (2017) showed increased indicine haplotypes in the bovine leucocyte antigen region of Brangus cattle raised in Argentina, potentially due to selection for adaptation to the environments. In a brief search of bovine QTL in the QTLdb database, we found two postnatal growth traits and one body mass QTL reported on these two chromosomes, which were located on chromosome 15 at 61.6 Mb (body weight gain) and 17.0 Mb (birth body weight) (Snelling et al., 2010), and chromosome 17 at 12.0 Mb (average daily gain) (Rolf et al., 2012). These two chromosomes had the highest Angus CBC (chromosome 15: 79.84%; chromosome 17: 79.21%). These QTL were also located in regions with high Angus proportions. There are QTLs associated with disease resistance on chromosome 5 (Machado et al., 2010), which had the highest Braham CBC (43.14%). There were also health-related QTLs on chromosome 5, which included infectious bovine keratoconjunctivitis susceptibility (Kizilkaya et al., 2013), cold tolerance (Howard et al., 2014), and immune capacity (Leach et al., 2010), to list a few of them. All these QTLs were located in regions with high Brahman proportions in chromosome 5. There was a pleiotropic QTL affecting birth, yearling, and mature weights on chromosome 20 at 7–8 Mb (Weng et al., 2016), which was a region with a high Angus proportion. However, this chromosome had a high Brahman proportion (40.2%) in our study and Paim T. D. P. et al. (2020). The presence of Brahman favorable alleles in chromosomal regions with high Angus breed proportions, or vice versa, exemplifies the successful complementary of favorable alleles for the traits of interest from both ancestral breeds.

Gene set enrichment analysis was conducted with functional genes and QTL extracted on chromosome segments with high ( ≥ 75 % ) Angus and high ( ≥ 50 % ) Brahman breed proportions, respectively, evaluated on 1-Mb chromosomal intervals (Figure 4). We adopted these two cutoff threshold values because each represented an equal (12.5%) upward deviation from their expected values. We computed SBC on 1-Mb intervals for the subsequent gene set enrichment analysis because we rendered it an appropriate window size to capture gene regions of interest. Based on the Bos taurus UMD 3.1.1 assembly, the genome size of domestic cattle is approximately 2,670 Mb, which contains 26,815 genes (https://www.ncbi.nlm.nih.gov/genome/annotation_euk/Bos_taurus/105/#FeatureCountsStats). Hence, the average spacing between neighboring genes is around 0.1 Mb. In contrast, the chromosome regions defined on 5-Mb or 10-Mb intervals could be too broad. Nevertheless, SNPs on 1 Mb intervals had higher LD than those on 5-Mb or 10-Mb intervals. The average R ² LD per segment on each chromosome ranged from 0.11 to 0.16 in 1-Mb intervals, from 0.06 to 0.10 in 5-Mb intervals, and from 0.05 to 0.08 in 10-Mb intervals (Table 2). When computing the likelihood, we noted that SNPs with higher LD tended to give more weight to these highly linked SNPs or their regions.

There were 852 segments with Angus SBC ≥75% and 169 segments with Brahman SBC ≥50%, which accounted for 33.8% and 6.7%, respectively, of the 2,522 1-Mb chromosomal segments on the genome (Table 7.). The total length of chromosomal segments with ≥ 75% Angus proportions was the longest on chromosome one and the shortest on chromosomes 5 and 13. Relatively speaking, chromosome 15 had the largest percentage (64.5%) of high Angus breed proportion segments, and Chromosome five had the least percentage (5.8%) of total high Angus breed proportion segments. Twenty-five chromosomes had more than 20% of the chromosome length as high ( ≥ 75 % ) Angus proportion regions. In contrast, only two chromosomes (5 and 20) had more than 20% of their total length as high ( ≥ 50 % ) Brahman breed proportions. Adjacent high Angus proportion segments often joined together to form larger blocks, presenting almost in all 29 chromosomes (Figure 4A). In contrast, high Brahman proportion segments were relatively sparse and isolated, though some formed small blocks, and they were present only on seven (2, 5, 6, 9, 10, 20, and 21) chromosomes (Figure 4B). These results agreed with a previous report yet with a different approach. Using a chromosome painting approach based on a copying model (Lawson et al., 2012), Paim T. D. P. et al. (2020) showed that chromosomal regions with high Angus breed proportions were prevalent on chromosomes, often large blocks, yet chromosome segments with high Brahman breed proportion were fewer and isolated. A follow-up study on selection signatures in Brangus revealed that the majority of selection signatures in Brangus cattle came from Angus (Paim T. D. P. et al., 2020). The copying model related the patterns of LD across chromosomes to the underlying recombination process and used a hidden Markov method to reconstruct a sampled haplotype. In the present study, we directly estimated breed compositions on each chromosome flanked by varying window sizes based on a mixture model. The admixture coefficients inferred from the admixture model can be probabilistically interpreted as reflecting that as identity-in-state (Jannink and Wu, 2003). Yet, when confined to long chromosomal chunks, it approximated the probability of identity by descent (Browning, 2008). Hence, though using a different approach, we came to similar findings.

In theory, crossing breaks and shuffles chromosomes randomly over time when directional selection is absent, which reflects genetic drift. However, with selection, it increases and even fixes favorable alleles. Meanwhile, it also changes the allelic frequencies of genes in LD with the genes under selection due to genomic hitchhiking, also known as genetic draft (Smith and Haigh, 1974; Ma et al., 2019). In other words, genomic hitchhiking occurs when a polymorphic locus is in LD with a second locus undergoing a selective sweep. As a result, the linked allele will also increase in frequency, in some cases, until it becomes fixed in the population. Overall, genomic hitchhiking reduces genetic variation and leaves footprints across the genome known as the signatures of selection (Sabeti et al., 2007; Singh et al., 2020). The many high Angus breed proportion regions reflected the presence of multiple favorable alleles scattered across the chromosomes. We observed the prevalence of large blocks with high Angus breed proportions (Figure 4A), possibly resulting from the genomic hitchhiking effects when selecting Brangus for Angus favorable traits. This is equivalent to saying that genomic hitchhiking effects were strong around the genomic regions with Angus favorable alleles, leading to the presence of large blocks with high Angus breed proportions. In contrast, genomic hitchhiking effects were weak around Brahman favorable alleles because chromosomal blocks with high Brahman breed proportions were relatively few, isolated, and small in size.

There were 9,025 genes on the chromosomal segments with Angus SBC ≥75% (Supplementary Table S1) and 1,877 genes on the chromosomal segments with Brahman SBC ≥50% (Supplementary Table S2). Many genes in high (≥75%) Angus regions are responsible for biological processes related to animal development, such as regulation of biological processes, anatomical structure development, anatomical structural morphogenesis, animal organ development, and skeletal system development, and, in KEGG, related to hormone regulation (such as the Estrogen signaling pathway). For example, the system development (GO:0048731) category’s related child terms include system development, such as central nervous, respiratory, and endocrine. We also found many genes associated with carcass and meat quality traits, such as PPP1R3B (Edwards et al., 2003; Cinar et al., 2012), ASXL1 (Grigoletto et al., 2020), DNMT3B (Liu et al., 2012), and TMEM68 (Lindholm-Perry et al., 2012; Terakado et al., 2018; Edea et al., 2020), just to list a few. Our gene list also included LEP on chromosome four and PLAG1 on chromosome 14. Paim T. D. P. et al. (2020) previously found these two genes in high Angus regions. The LEP gene is expressed in adipose tissue and codes for leptin, a hormone known to regulate feed intake and energy balance in mammals (Woronuk et al., 2012). This gene is associated with marbling, fat thickness, rib eye area, and feed intake in several beef cattle breeds (Souza et al., 2010; Woronuk et al., 2012; Kononoff et al., 2017). Leptin is an essential gene for puberty onset (Williams et al., 2002). This gene could be inherited from Angus ancestors, or its frequency was increased by the selection of Brangus for early puberty since breed formation because Bos indicus heifers have challenges achieving puberty early in life (Sartori et al., 2010; Fortes et al., 2012). PLAG1 is involved in regulating stature and weight (Littlejohn et al., 2012; Pryce et al., 2012; Song et al., 2016). This gene is associated with yearling weight in Australian Tropical Composite breeds (Porto-Neto et al., 2014). There is still another gene, XKR4, which is close to PLAG1. The XKR4 gene is associated with subcutaneous rump fat thickness, scrotal circumference, serum concentration of prolactin, and sexual precocity (Fortes et al., 2012; Porto Neto et al., 2012; Bastin et al., 2014; Takada et al., 2018). The genes in the high ( ≥ 50 % ) Brahman regions are primarily responsible for biosynthetic-related biological processes (e.g., cytolysis), molecular biological functions related to enzyme activity (e.g., lysozyme activity, peptidoglycan muralytic activity, hydrolase activity, and serine-type endopeptidase activity), and diseases and immunity (e.g., MHC class II protein complex, MHC protein complex, type I diabetes mellitus, allograft rejection, graft-versus-host disease, autoimmune thyroid disease, and pathogenic Escherichia coli infection). For example, peptidoglycan muralytic activity (GO:0061783), which contributes to the degradation of peptidoglycan, is a major structural component of bacterial cell walls (Nelson et al., 2012); Another example is the MHC Class II protein complex (GO:0042613). MHC is involved in the immune process and plays the role of transmitting antigens (Rodgers and Cook, 2005). Still, the gene functionalities in chromosomal regions with high Angus breed proportions were diverse, possibly due to genomic hitchhiking of genes linked to the favorable alleles under selection. For example, there was a category of genes called “cellular response to stress”. Cells respond to stress in various ways, from activating pathways that promote survival to eliciting programmed cell death that eliminates damaged cells. Cell death research has attracted much attention in the last 2 decades also because of its relevance to development, degenerative diseases, and cancer in human (Fulda et al., 2010). Overall, the Brangus cattle have successfully combined the favorable traits of the two highly successful parent breeds.

QTLs were extracted from the chromosomal regions with the top 1% highest Angus (Angus SBC ≥ 91.1 % ) and Brahman (Braham SBC ≥ 59.4 % ) breed proportions, respectively (Supplementary Table S3). The QTLs in the top 1% Angus regions are related to meat, carcass, production, and reproduction. This list included, for example, birth weight QTL (chromosomes three and 8), weaning weight QTL (chromosomes 1, 3, 15, 18, and 23), yearling weight QTL (chromosomes 3 and 10), mature body weight QTL (chromosomes 3, 4, 10, 17, and 18), carcass weight QTL (chromosomes three and 8), marbling score QTL (chromosomes 15, 18, and 22), fat thickness QTL (chromosomes 3, 4, 17, 18, and 23), calving ease QTL (chromosomes 3, 8, 10, and 17), and QTL for Longissimus muscle area (chromosomes 3, 8, 17, and 18). Some QTL have pleiotropic effects. The QTL in the regions with 1% highest Brahman proportions are related to health, such as bovine respiratory disease susceptibility (chromosome 2: 117.3–134.4 Mb), insulin-like growth factor 1 level on (chromosome 5: 40.3–59.7 Mb; chromosome 14: 26.5–27.0 Mb), and blood cortisol level (chromosome 16: 29.7 Mb).