Applications of livestock monitoring devices and machine learning algorithms in animal production and reproduction: an overview.

Some sectors of animal production and reproduction have shown great technological advances due to the development of research areas such as Precision Livestock Farming (PLF). PLF is an innovative approach that allows animals to be monitored, through the adoption of cutting-edge technologies that continuously collect real-time data by combining the use of sensors with advanced algorithms to provide decision tools for farmers. Artificial Intelligence (AI) is a field that merges computer science and large datasets to create expert systems that are able to generate predictions and classifications similarly to human intelligence. In a simplified manner, Machine Learning (ML) is a branch of AI, and can be considered as a broader field that encompasses Deep Learning (DL, a Neural Network formed by at least three layers), generating a hierarchy of subsets formed by AI, ML and DL, respectively. Both ML and DL provide innovative methods for analyzing data, especially beneficial for large datasets commonly found in livestock-related activities. These approaches enable the extraction of valuable insights to address issues related to behavior, health, reproduction, production, and the environment, facilitating informed decision-making. In order to create the referred technologies, studies generally go through five steps involving data processing: acquisition, transferring, storage, analysis and delivery of results. Although the data collection and analysis steps are usually thoroughly reported by the scientific community, a good execution of each step is essential to achieve good and credible results, which impacts the degree of acceptance of the proposed technologies in real life practical circumstances. In this context, the present work aims to describe an overview of the current implementations of ML/DL in livestock reproduction and production, as well to identify potential challenges and critical points in each of the five steps mentioned, which can affect results and application of AI techniques by farmers in practical situations.

A Random Forest-Based Genome-Wide Scan Reveals Fertility-Related Candidate Genes and Potential Inter-Chromosomal Epistatic Regions Associated With Age at First Calving in Nellore Cattle.

This study aimed to perform a genome-wide association analysis (GWAS) using the Random Forest (RF) approach for scanning candidate genes for age at first calving (AFC) in Nellore cattle. Additionally, potential epistatic effects were investigated using linear mixed models with pairwise interactions between all markers with high importance scores within the tree ensemble non-linear structure. Data from Nellore cattle were used, including records of animals born between 1984 and 2015 and raised in commercial herds located in different regions of Brazil. The estimated breeding values (EBV) were computed and used as the response variable in the genomic analyses. After quality control, the remaining number of animals and SNPs considered were 3,174 and 360,130, respectively. Five independent RF analyses were carried out, considering different initialization seeds. The importance score of each SNP was averaged across the independent RF analyses to rank the markers according to their predictive relevance. A total of 117 SNPs associated with AFC were identified, which spanned 10 autosomes (2, 3, 5, 10, 11, 17, 18, 21, 24, and 25). In total, 23 non-overlapping genomic regions embedded 262 candidate genes for AFC. Enrichment analysis and previous evidence in the literature revealed that many candidate genes annotated close to the lead SNPs have key roles in fertility, including embryo pre-implantation and development, embryonic viability, male germinal cell maturation, and pheromone recognition. Furthermore, some genomic regions previously associated with fertility and growth traits in Nellore cattle were also detected in the present study, reinforcing the effectiveness of RF for pre-screening candidate regions associated with complex traits. Complementary analyses revealed that many SNPs top-ranked in the RF-based GWAS did not present a strong marginal linear effect but are potentially involved in epistatic hotspots between genomic regions in different autosomes, remarkably in the BTAs 3, 5, 11, and 21. The reported results are expected to enhance the understanding of genetic mechanisms involved in the biological regulation of AFC in this cattle breed.

Practical implications of using non-relational databases to store large genomic data files and novel phenotypes.

The objective of our study was to provide practical directions on the storage of genomic information and novel phenotypes (treated here as unstructured data) using a non-relational database. The MongoDB technology was assessed for this purpose, enabling frequent data transactions involving numerous individuals under genetic evaluation. Our study investigated different genomic (Illumina Final Report, PLINK, 0125, FASTQ, and VCF formats) and phenotypic (including media files) information, using both real and simulated datasets. Advantages of our centralized database concept include the sublinear running time for queries after increasing the number of samples/markers exponentially, in addition to the comprehensive management of distinct data formats while searching for specific genomic regions. A comparison of our non-relational and generic solution, with an existing relational approach (developed for tabular data types using 2 bits to store genotypes), showed reduced importing time to handle 50M SNPs (PLINK format) achieved by the relational schema. Our experimental results also reinforce that data conversion is a costly step required to manage genomic data into both relational and non-relational database systems, and therefore, must be carefully treated for large applications.

Reduced rank analysis of morphometric and functional traits in Campolina horses.

Multitrait models can increase the accuracy of breeding value prediction and reduce bias due to selection by using traits measured before and after it has occurred. However, as the number of traits grows, a similar trend is expected for the number of parameters to be estimated, which directly affects the computing power and the amount of data required. The aim of the present study was to apply reduced rank (principal components model-PCM) and factor analytical models (FAM), to estimate (co)variance components for nineteen traits, jointly evaluated in a single analysis in Campolina horses. A total of 18 morphometric traits (MT) and one gait visual score (GtS), along with genealogical records of 48,806 horses, were analysed under a restricted maximum likelihood framework. Nine PCM, nine FAM and one standard multitrait model (MTM) were fitted to the data and compared to find the best suitable model. Based on Bayesian information criterion, the best model was the FAM option, considering five common factors (FAM5). After performing an intraclass analysis, none of MT were genetically negatively correlated, whereas GtS was negatively related to all MT, except for the genetic correlations among GtS and BLL, and between GtS and BLLBL (0.01 and 0.10 respectively). From all MT, two traits were derived computing ratios involving other traits, those had negative correlations with others MT, but all favourable for selection. Similar patterns were observed between the genetic parameters obtained from MTM and FAM5 respectively. The heritability estimates ranged from 0.09 (head width) to 0.47 (height at withers). Our results indicated that FAM was efficient to reduce the multitrait analysis dimensionality, and therefore, traits can be combined based on the first three eigenvectors from the additive genetic (co)variance matrix. In addition, there was sufficient genetic variation for selection, benefiting its potential implementation in a breeding program.

Publisher Correction: Uncovering Sub-Structure and Genomic Profiles in Across-Countries Subpopulations of Angus Cattle.

An amendment to this paper has been published and can be accessed via a link at the top of the paper.

Whole-genome sequencing provides new insights into genetic mechanisms of tropical adaptation in Nellore (Bos primigenius indicus).

Most of the knowledge about genetic variants at the sequence level in cattle is for Bos primigenius taurus populations. Here, we presented a complete genomic characterization of 52 Nellore (Bos primigenius indicus) bulls, revealing specific zebu DNA variants with putative impact in tropical adaptation and productive traits. Single nucleotide polymorphisms (SNPs) and insertion/deletion (INDELs) mutations were identified using the newest bovine reference genome ARS_UCD1.2, and variant functional consequences were predicted using the Ensembl VEP software. A total of 35,753,707 SNPs and 4,492,636 INDELs were detected and annotated to their functional effects. We identified 400 genes that comprised both, a SNP and an INDEL, of high functional impact on proteins (i.e. variants that cause protein truncation, loss of function or triggering nonsense-mediated decay). Among these, we highlight the following genes: BoLA, associated with cattle immune response to infections and reproduction aspects; HSPA8, DNAJC27, and DNAJC28, involved with thermoregulatory protective mechanisms in mammals; and many olfactory signaling pathway related genes that are important genetic factors in the evolution of mammalian species. All these functional aspects are directly related to cattle adaptability to tropical environments.

Uncovering Sub-Structure and Genomic Profiles in Across-Countries Subpopulations of Angus Cattle.

Highlighting genomic profiles for geographically distinct subpopulations of the same breed may provide insights into adaptation mechanisms to different environments, reveal genomic regions divergently selected, and offer initial guidance to joint genomic analysis. Here, we characterized similarities and differences between the genomic patterns of Angus subpopulations, born and raised in Canada (N = 382) and Brazil (N = 566). Furthermore, we systematically scanned for selection signatures based on the detection of autozygosity islands common between the two subpopulations, and signals of divergent selection, via FST and varLD tests. The principal component analysis revealed a sub-structure with a close connection between the two subpopulations. The averages of genomic relationships, inbreeding coefficients, and linkage disequilibrium at varying genomic distances were rather similar across them, suggesting non-accentuated differences in overall genomic diversity. Autozygosity islands revealed selection signatures common to both subpopulations at chromosomes 13 (63.77-65.25 Mb) and 14 (22.81-23.57 Mb), which are notably known regions affecting growth traits. Nevertheless, further autozygosity islands along with FST and varLD tests unravel particular sites with accentuated population subdivision at BTAs 7 and 18 overlapping with known QTL and candidate genes of reproductive performance, thermoregulation, and resistance to infectious diseases. Our findings indicate overall genomic similarity between Angus subpopulations, with noticeable signals of divergent selection in genomic regions associated with the adaptation in different environments.

Genetic analysis of in-vitro embryo production traits in Dairy Gir cattle.

The potential of dams as oocyte donors can be a selection criterion for animal breeding programs, but also an involuntary driver of the process. In both cases, it is important to determine genetic components influencing the outcome of in vitro embryo production (IVEP). The objective of the present study was to perform a detailed genetic analysis for in vitro embryo production traits in Dairy Gir cows. A dataset containing 11,450 records of ovum pick-up (OPU) and in vitro fertilization (IVF) procedures from 2684 Dairy Gir donors was evaluated. Analyzed traits were number (NOV) and percentage (POV) of viable oocytes; number (NGI) and percentage (PGI) of grade I oocytes; number (NEMB) and percentage (PEMB) of viable embryos. All analyzes were performed using animal models by a Bayesian framework. Heritability estimates varied from 0.16 to 0.32 for count traits and from 0.01 to 0.06 for percentage traits. The proportion of the total variation represented by the additive genetic effect of sire (semen used in IVF) for NEMB and PEMB was 7% and 5% respectively. Associations between estimated breeding values from progeny tested bulls for IVEP traits, milk production, age at first calving and conformation traits were mainly low or close to zero. Results indicate that selection for IVEP traits is possible in Dairy Gir cattle and would not impair genetic progress for traits already considered as selection criteria. The NOV seems to be a promising target trait. However, a selection index could help to avoid the use of sires with negative genetic merit for percentage traits, minimizing possible deterioration in the long term.

Assessment of runs of homozygosity islands and estimates of genomic inbreeding in Gyr (Bos indicus) dairy cattle.

BACKGROUND: Runs of homozygosity (ROH) are continuous homozygous segments of the DNA sequence. They have been applied to quantify individual autozygosity and used as a potential inbreeding measure in livestock species. The aim of the present study was (i) to investigate genome-wide autozygosity to identify and characterize ROH patterns in Gyr dairy cattle genome; (ii) identify ROH islands for gene content and enrichment in segments shared by more than 50% of the samples, and (iii) compare estimates of molecular inbreeding calculated from ROH (FROH), genomic relationship matrix approach (FGRM) and based on the observed versus expected number of homozygous genotypes (FHOM), and from pedigree-based coefficient (FPED). RESULTS: ROH were identified in all animals, with an average number of 55.12 ± 10.37 segments and a mean length of 3.17 Mb. Short segments (ROH1-2 Mb) were abundant through the genomes, which accounted for 60% of all segments identified, even though the proportion of the genome covered by them was relatively small. The findings obtained in this study suggest that on average 7.01% (175.28 Mb) of the genome of this population is autozygous. Overlapping ROH were evident across the genomes and 14 regions were identified with ROH frequencies exceeding 50% of the whole population. Genes associated with lactation (TRAPPC9), milk yield and composition (IRS2 and ANG), and heat adaptation (HSF1, HSPB1, and HSPE1), were identified. Inbreeding coefficients were estimated through the application of FROH, FGRM, FHOM, and FPED approaches. FPED estimates ranged from 0.00 to 0.327 and FROH from 0.001 to 0.201. Low to moderate correlations were observed between FPED-FROH and FGRM-FROH, with values ranging from -0.11 to 0.51. Low to high correlations were observed between FROH-FHOM and moderate between FPED-FHOM and FGRM-FHOM. Correlations between FROH from different lengths and FPED gradually increased with ROH length. CONCLUSIONS: Genes inside ROH islands suggest a strong selection for dairy traits and enrichment for Gyr cattle environmental adaptation. Furthermore, low FPED-FROH correlations for small segments indicate that FPED estimates are not the most suitable method to capture ancient inbreeding. The existence of a moderate correlation between larger ROH indicates that FROH can be used as an alternative to inbreeding estimates in the absence of pedigree records.

Candidate genes for male and female reproductive traits in Canchim beef cattle.

BACKGROUND: Beef cattle breeding programs in Brazil have placed greater emphasis on the genomic study of reproductive traits of males and females due to their economic importance. In this study, genome-wide associations were assessed for scrotal circumference at 210 d of age, scrotal circumference at 420 d of age, age at first calving, and age at second calving, in Canchim beef cattle. Data quality control was conducted resulting in 672,778 SNPs and 392 animals. RESULTS: Associated SNPs were observed for scrotal circumference at 420 d of age (435 SNPs), followed by scrotal circumference at 210 d of age (12 SNPs), age at first calving (six SNPs), and age at second calving (four SNPs). We investigated whether significant SNPs were within genic or surrounding regions. Biological processes of genes were associated with immune system, multicellular organismal process, response to stimulus, apoptotic process, cellular component organization or biogenesis, biological adhesion, and reproduction. CONCLUSIONS: Few associations were observed for scrotal circumference at 210 d of age, age at first calving, and age at second calving, reinforcing their polygenic inheritance and the complexity of understanding the genetic architecture of reproductive traits. Finding many associations for scrotal circumference at 420 d of age in various regions of the Canchim genome also reveals the difficulty of targeting specific candidate genes that could act on fertility; nonetheless, the high linkage disequilibrium between loci herein estimated could aid to overcome this issue. Therefore, all relevant information about genomic regions influencing reproductive traits may contribute to target candidate genes for further investigation of causal mutations and aid in future genomic studies in Canchim cattle to improve the breeding program.

Using a system of differential equations that models cattle growth to uncover the genetic basis of complex traits.

The interplay between dynamic models of biological systems and genomics is based on the assumption that genetic variation of the complex trait (i.e., outcome of model behavior) arises from component traits (i.e., model parameters) in lower hierarchical levels. In order to provide a proof of concept of this statement for a cattle growth model, we ask whether model parameters map genomic regions that harbor quantitative trait loci (QTLs) already described for the complex trait. We conducted a genome-wide association study (GWAS) with a Bayesian hierarchical LASSO method in two parameters of the Davis Growth Model, a system of three ordinary differential equations describing DNA accretion, protein synthesis and degradation, and fat synthesis. Phenotypic and genotypic data were available for 893 Nellore (Bos indicus) cattle. Computed values for parameter k1 (DNA accretion rate) ranged from 0.005 ± 0.003 and for α (constant for energy for maintenance requirement) 0.134 ± 0.024. The expected biological interpretation of the parameters is confirmed by QTLs mapped for k1 and α. QTLs within genomic regions mapped for k1 are expected to be correlated with the DNA pool: body size and weight. Single nucleotide polymorphisms (SNPs) which were significant for α mapped QTLs that had already been associated with residual feed intake, feed conversion ratio, average daily gain (ADG), body weight, and also dry matter intake. SNPs identified for k1 were able to additionally explain 2.2% of the phenotypic variability of the complex ADG, even when SNPs for k1 did not match the genomic regions associated with ADG. Although improvements are needed, our findings suggest that genomic analysis on component traits may help to uncover the genetic basis of more complex traits, particularly when lower biological hierarchies are mechanistically described by mathematical simulation models.

Study on the introgression of beef breeds in Canchim cattle using single nucleotide polymorphism markers.

The aim of this study was to evaluate the level of introgression of breeds in the Canchim (CA: 62.5% Charolais-37.5% Zebu) and MA genetic group (MA: 65.6% Charolais-34.4% Zebu) cattle using genomic information on Charolais (CH), Nelore (NE), and Indubrasil (IB) breeds. The number of animals used was 395 (CA and MA), 763 (NE), 338 (CH), and 37 (IB). The Bovine50SNP BeadChip from Illumina panel was used to estimate the levels of introgression of breeds considering the Maximum likelihood, Bayesian, and Single Regression method. After genotype quality control, 32,308 SNPs were considered in the analysis. Furthermore, three thresholds to prune out SNPs in linkage disequilibrium higher than 0.10, 0.05, and 0.01 were considered, resulting in 15,286, 7,652, and 1,582 SNPs, respectively. For k = 2, the proportion of taurine and indicine varied from the expected proportion based on pedigree for all methods studied. For k = 3, the Regression method was able to differentiate the animals in three main clusters assigned to each purebred breed, showing more reasonable according to its biological viewpoint. Analyzing the data considering k = 2 seems to be more appropriate for Canchim-MA animals due to its biological interpretation. The usage of 32,308 SNPs in the analyses resulted in similar findings between the estimated and expected breed proportions. Using the Regression approach, a contribution of Indubrasil was observed in Canchim-MA when k = 3 was considered. Genetic parameter estimation could account for this breed composition information as a source of variation in order to improve the accuracy of genetic models. Our findings may help assemble appropriate reference populations for genomic prediction for Canchim-MA in order to improve prediction accuracy. Using the information on the level of introgression in each individual could also be useful in breeding or crossing design to improve individual heterosis in crossbred cattle.

Identification of Candidate Genes for Reactivity in Guzerat (Bos indicus) Cattle: A Genome-Wide Association Study.

Temperament is fundamental to animal production due to its direct influence on the animal-herdsman relationship. When compared to calm animals, the aggressive, anxious or fearful ones exhibit less weight gain, lower reproductive efficiency, decreased milk production and higher herd maintenance costs, all of which contribute to reduced profits. However, temperament is a trait that is complex and difficult to assess. Recently, a new quantitative system, REATEST®, for assessing reactivity, a phenotype of temperament, was developed. Herein, we describe the results of a Genome-wide association study for reactivity, assessed using REATEST® with a sample of 754 females from five dual-purpose (milk and meat production) Guzerat (Bos indicus) herds. Genotyping was performed using a 50k SNP chip and a two-step mixed model approach (Grammar-Gamma) with a one-by-one marker regression was used to identify QTLs. QTLs for reactivity were identified on chromosomes BTA1, BTA5, BTA14, and BTA25. Five intronic and two intergenic markers were significantly associated with reactivity. POU1F1, DRD3, VWA3A, ZBTB20, EPHA6, SNRPF and NTN4 were identified as candidate genes. Previous QTL reports for temperament traits, covering areas surrounding the SNPs/genes identified here, further corroborate these associations. The seven genes identified in the present study explain 20.5% of reactivity variance and give a better understanding of temperament biology.

Copy number variations and genome-wide associations reveal putative genes and metabolic pathways involved with the feed conversion ratio in beef cattle.

The use of genome-wide association results combined with other genomic approaches may uncover genes and metabolic pathways related to complex traits. In this study, the phenotypic and genotypic data of 1475 Nellore (Bos indicus) cattle and 941,033 single nucleotide polymorphisms (SNPs) were used for genome-wide association study (GWAS) and copy number variations (CNVs) analysis in order to identify candidate genes and putative pathways involved with the feed conversion ratio (FCR). The GWAS was based on the Bayes B approach analyzing genomic windows with multiple regression models to estimate the proportion of genetic variance explained by each window. The CNVs were detected with PennCNV software using the log R ratio and B allele frequency data. CNV regions (CNVRs) were identified with CNVRuler and a linear regression was used to associate CNVRs and the FCR. Functional annotation of associated genomic regions was performed with the Database for Annotation, Visualization and Integrated Discovery (DAVID) and the metabolic pathways were obtained from the Kyoto Encyclopedia of Genes and Genomes (KEGG). We showed five genomic windows distributed over chromosomes 4, 6, 7, 8, and 24 that explain 12 % of the total genetic variance for FCR, and detected 12 CNVRs (chromosomes 1, 5, 7, 10, and 12) significantly associated [false discovery rate (FDR) < 0.05] with the FCR. Significant genomic regions (GWAS and CNV) harbor candidate genes involved in pathways related to energetic, lipid, and protein metabolism. The metabolic pathways found in this study are related to processes directly connected to feed efficiency in beef cattle. It was observed that, even though different genomic regions and genes were found between the two approaches (GWAS and CNV), the metabolic processes covered were related to each other. Therefore, a combination of the approaches complement each other and lead to a better understanding of the FCR.

Linkage disequilibrium and haplotype block structure in a composite beef cattle breed.

BACKGROUND: The development of linkage disequilibrium (LD) maps and the characterization of haplotype block structure at the population level are useful parameters for guiding genome wide association (GWA) studies, and for understanding the nature of non-linear association between phenotypes and genes. The elucidation of haplotype block structure can reduce the information of several single nucleotide polymorphisms (SNP) into the information of a haplotype block, reducing the number of SNPs in a coherent way for consideration in GWA and genomic selection studies. RESULTS: The maximum average LD, measured by r2 varied between 0.33 to 0.40 at a distance of < 2.5 kb, and the minimum average values of r2 varied between 0.05 to 0.07 at distances ranging from 400 to 500 kb, clearly showing that the average r2 reduced with the increase in SNP pair distances. The persistence of LD phase showed higher values at shorter genomic distances, decreasing with the increase in physical distance, varying from 0.96 at a distance of < 2.5 kb to 0.66 at a distance from 400 to 500 kb. A total of 78% of all SNPs were clustered into haplotype blocks, covering 1,57 Mb of the total autosomal genome size. CONCLUSIONS: This study presented the first high density linkage disequilibrium map and haplotype block structure for a composite beef cattle population, and indicates that the high density SNP panel over 700 k can be used for genomic selection implementation and GWA studies for Canchim beef cattle.