Livestock breeding programs increasingly seek to improve animal productivity while curbing input expenditures and resource usage, bringing feed intake and enteric methane emissions to the forefront of modern breeding objectives. Because individual phenotypic recordings for these sustainability metrics have expanded across sheep and cattle populations, animal geneticists face the practical challenge of integrating resource efficiency into selection programs. In a theoretical analysis published in Genetics Selection Evolution, researcher Julius H. J. van der Werf applied classical selection index theory to examine whether incorporating efficiency ratios directly into multi-trait selection schemes offers tangible genetic advantages. The investigation evaluated single-trait selection for ratio indicators, compared them with linear selection indices composed of individual component traits, and analyzed combined index configurations that paired a ratio trait with one of its underlying component variables. The central mathematical finding demonstrates that an index incorporating a ratio trait alongside at least one underlying component trait produces genetic gains identical to a conventional linear index containing both component traits, provided that economic weights are appropriately adjusted to account for the mathematical interdependencies between the variables.
Animal breeders have long targeted feed efficiency to minimize costly feed inputs relative to total product output, a practice that originated in intensive livestock industries with automated feed recording systems. In ruminant production systems, measuring feed consumption has acquired critical environmental significance alongside financial profitability because enteric methane synthesis is directly tied to the volume of feed ingested. Researchers have frequently proposed using simple numerical quotients as direct breeding criteria, including feed conversion ratio, defined as feed intake divided by production, or feed efficiency, defined as product output divided by feed intake. Analogous ratio formulations exist for greenhouse gas mitigation, including methane intensity, which calculates enteric methane output per unit of animal product, and methane yield, which standardizes methane production per unit of dry matter consumed. Although expressing complex biological relationships as single quotient values appears intuitively straightforward to industry practitioners, previous statistical investigations identified critical drawbacks in ratio selection. Specifically, single-trait selection focused solely on an efficiency ratio implicitly imposes economic emphasis proportional to phenotypic trait means rather than their true financial values, frequently causing suboptimal overall selection responses.
To rigorously predict how selection acts upon quotient traits, the analytical framework employed a first-order Taylor series approximation formulated by Lynch and Walsh to approximate phenotypic and additive genetic variance terms from the underlying component traits. For any ratio between two phenotypic measurements, designated as variable x in the numerator and variable y in the denominator with corresponding means and variances, the overall phenotypic variance depends directly on the squared ratio of trait means, the coefficient of phenotypic variation for each component trait, and the phenotypic correlation linking both variables. When the phenotypic correlation between numerator and denominator measurements is lower, the resulting phenotypic variance of the calculated ratio trait becomes progressively larger. In a completely parallel manner, the additive genetic variance of the ratio trait scales with the squared ratio of trait averages, the coefficients of genetic variation belonging to the individual component traits, and their genetic correlation. The derivation establishes that the genetic correlation between component traits exerts no influence on the phenotypic variance of the ratio, while the phenotypic correlation does not alter the underlying additive genetic variance.
The heritability of a ratio trait is mathematically derived as the proportion of its additive genetic variance divided by its total phenotypic variance, simplifying into an expression governed by the ratio of genetic coefficients of variation, the ratio of phenotypic coefficients of variation, the component trait heritabilities, and the correlation coefficients. Trait means cancel out of this formulation entirely, demonstrating that heritability estimates for efficiency ratios remain completely independent of the arbitrary physical units of measurement used for the component variables. When the component traits possess equal heritabilities and identical genetic and phenotypic correlations, the heritability of the resulting ratio exactly equals that of its components. However, when the phenotypic correlation exceeds the genetic correlation between the component traits, phenotypic variance is reduced to a greater degree than genetic variance, creating a noticeably higher heritability for the ratio trait. Conversely, when the genetic correlation substantially exceeds the environmental correlation, the ratio trait exhibits very low heritability, whereas negative differences between genetic and environmental correlations elevate the ratio trait heritability to higher values.
Approximating the phenotypic and genetic correlations connecting a ratio trait to its original component traits reveals the directional mechanics of artificial selection. Using Pearson mathematical approximations, the genetic correlation between a ratio and its numerator trait is generally positive, whereas its genetic correlation with the denominator trait is predominantly negative. When coefficients of genetic variation and heritabilities are identical between the component variables, the ratio of coefficients of genetic variation equals one, causing the genetic correlation between the ratio and the numerator to precisely mirror the magnitude of its genetic correlation with the denominator while carrying an opposite mathematical sign. When the denominator trait exhibits a larger coefficient of genetic variation, the parameter ratio drops below one, characterizing a denominator-dominant trait where the ratio correlates much more strongly with the denominator. In contrast, when the numerator exhibits greater genetic variability, the ratio exceeds one, creating numerator dominance. In extreme cases where component genetic variations diverge markedly, the genetic correlation of the ratio with both component traits can simultaneously become positive or negative.
When animal breeders perform single-trait selection targeting a lower ratio value, the expected genetic response across the underlying component traits is dictated by the ratio of their genetic correlations with that quotient. For individual phenotypic mass selection or single-trait estimated breeding value selection, the absolute correlated response per unit of genetic standard deviation is equivalent between components only when their coefficients of genetic variation are perfectly matched. If the denominator trait displays greater relative genetic variation, selection aimed at reducing the overall quotient drives a substantial positive genetic change in the denominator, while exerting only minor, negligible shifts upon the numerator. Furthermore, if the ratio of genetic coefficients of variation falls below the genetic correlation between components, the numerator can inadvertently shift in an unfavorable positive direction. Because single-trait selection responses are governed strictly by relative genetic coefficients of variation rather than relative economic returns, pursuing genetic improvement solely through an unadjusted efficiency quotient invariably produces an economically suboptimal selection trajectory.
Integrating efficiency quotients into multi-trait breeding programs requires reformulating the underlying selection index matrices and linear profit functions to accommodate the mathematical dependencies between correlated variables. Selection index theory demonstrates that a single-trait breeding objective targeting a ratio is equivalent to an unweighted linear objective where the component traits are assigned economic weights corresponding directly to the trait means of the opposite component. In more complex multi-trait breeding objectives where a ratio trait is incorporated alongside one of its own component traits, the economic weights must be mathematically adjusted to avoid double-counting overlapping biological variation. If a ratio trait is combined with the denominator trait in the breeding objective, the economic value assigned to the ratio trait must equal the mean of the denominator multiplied by the economic value of the numerator. Simultaneously, the economic weight assigned to the denominator trait must be modified to absorb the correlated economic contribution generated by corresponding shifts in the numerator trait, ensuring total equivalence to a linear multi-trait objective.
To validate the theoretical derivations with realistic agricultural parameters, the study evaluated a numerical case study of Australian Angus beef cattle measured in feedlot environments between five hundred and six hundred days of age. The empirical parameters, derived from published livestock evaluations by Torres-Vázquez and colleagues, examined feed intake as the numerator trait, average daily gain as the denominator trait, and feed conversion ratio as the calculated quotient. Economic values were established at negative sixty dollars per kilogram per day for feed intake to represent feed costs and positive seven hundred and fifty dollars per kilogram per day for average daily gain. The empirical feed intake mean was fourteen point nine kilograms, average daily gain averaged one point fifty-nine kilograms per day, and the genetic correlation connecting the two metrics was zero point seventy-two. Because average daily gain displayed a higher coefficient of genetic variation, the resulting parameter ratio equaled zero point seventy-six, confirming that feed conversion ratio behaves as a denominator-dominant trait in beef cattle populations.
Comparing multi-trait selection scenarios executed through the MTINDEX framework illustrated stark contrasts in economic efficiency across diverse index configurations. Single-trait selection aimed strictly at reducing the feed conversion ratio achieved genetic changes in units of genetic standard deviation of negative zero point zero four for feed intake and positive zero point thirty-eight for average daily gain. Although these correlated responses moved both biological traits in favorable directions by lowering consumption while increasing growth, the total financial return was seven point four percent lower than the optimal economic response achieved by a linear index comprising both component traits directly. The linear index generated superior economic returns by aggressively boosting average daily gain while accepting a slight correlated increase in feed intake, capitalizing on the substantially higher economic value of daily gain. However, when the feed conversion ratio was incorporated into a multi-trait index alongside either average daily gain or feed intake, the resulting selection response perfectly matched the optimal economic gain produced by the linear component index.
Beyond confirming mathematical equivalences, the findings highlight practical and statistical considerations for implementing efficiency and environmental traits in national livestock evaluations. A primary logistical obstacle to using ratio traits directly as selection criteria is the rigid requirement that both underlying phenotypes must be recorded simultaneously on the exact same animals. In contemporary genomic selection and single-step genetic evaluation programs, reference populations frequently record individual component traits across separate animal cohorts, such as measuring dry matter intake in specialized research herds while recording growth metrics across commercial herds. Linear multi-trait frameworks seamlessly evaluate individuals possessing single-trait records or missing data, whereas quotient metrics cannot be calculated directly from incomplete phenotypic pairs. Similar analytical complexities arise when evaluating methane intensity, which represents enteric emissions divided by product yield, because total meat output per breeding female reflects multiple contributing traits including reproductive rates, weaning percentages, and product quality. The study concludes that multi-trait genetic evaluation systems based directly on component traits offer superior statistical flexibility, naturally accommodate missing records, and permit transparent derivation of efficiency metrics post-evaluation.
Subject of Research: Theoretical evaluation of ratio traits in multiple-trait livestock selection indices
Article Title: Including ratio traits in multiple trait selection
Article References: van der Werf, J. H. J. (2026). Including ratio traits in multiple trait selection. Genetics Selection Evolution, 58(1), Article 87. https://doi.org/10.1186/s12711-026-01111-3
Image Credits: AI Generated
DOI: 10.1186/s12711-026-01111-3
Keywords: selection index theory, ratio traits, feed efficiency, enteric methane emissions, feed intake, average daily gain, feed conversion ratio, methane intensity, economic weights, coefficient of genetic variation, breeding objectives, genetic correlation, phenotypic correlation, livestock breeding, multiple trait selection, genomic selection
News Source: Juliet Wilcox. (November 12, 2026). Ratio Traits Match Linear Selection Indices When Paired With Component Traits. Scienmag.
