biomarkers

Abstract

The disclosure concerns a method for predicting the productivity of a milk- producing animal, a method of using a trained machine learning algorithm model to characterise a milk-producing animal as being of high productivity or low productivity, and a method of training a machine learning algorithm model accordingly. The disclosure also concerns a method of selecting a milk-producing animal for breeding. The disclosure further concerns a related computer program and data processing apparatus.

Description

[0001] BIOMARKERS

[0002] FIELD OF THE DISCLOSURE

[0003] The disclosure concerns a method for predicting the productivity of a milkproducing animal, a method of using a trained machine learning algorithm model to characterise a milk-producing animal as being of high productivity or low productivity, and a method of training a machine learning algorithm model accordingly. The disclosure also concerns a method of selecting a milk-producing animal for breeding. The disclosure further concerns a related computer program and data processing apparatus.

[0004] BACKGROUND

[0005] For many milk-producing farmers, it is important to maximise the productivity of their herd. Margins in farming are tight, and sub-optimal productivity can have great financial implications and affect the viability of a farm. Furthermore, productivity impacts the ability to farm sustainably. Milk-producing animals that have poor productivity, yet consume valuable resources and emit methane and carbon dioxide, have a negative effect on the sustainability of a farm. The ability to predict the productivity of an individual animal with a view to retaining high-performance individuals in the herd and / or selecting them for breeding is therefore desirable.

[0006] SUMMARY OF THE DISCLOSURE

[0007] The present inventors have discovered that the productivity of a milk-producing animal, such as a bovine, can be predicted from the level of expression of certain genes. In particular, they have identified that a lower expression of one or more haemoglobin genes such as HBA, HBA1, HBB, HBM and / or HBQ1 is associated with higher productivity. For instance, milk yield, milk fat percentage, milk protein percentage and / or milking speed may be improved in individuals with lower expression of HBA, HBA1, HBB, HBM and / or HBQ1, or their offspring. Accordingly, the productivity of a milk-producing animal can be predicted by interrogating expression of HBA, HBA1, HBB, HBM and / or HBQ1 in a sample obtained from the milk-producing animal, or a parent of the milk-producing animal. The predicted productivity can be used to determine how the milk-producing animal or parent is managed. For instance, a high productivity milk-producing animal may be retained in the herd, or a low productivity milk-producing animal may be removed from the herd, with a view of improving productivity at the herd level. Productivity of the herd may also be improved by selecting high productivity animals, or a parent giving rise to high productivity offspring, for use in breeding new herd members. Improvements in productivity have important implications for sustainable farming and the economic viability of the herd.

[0008] The present inventors have further developed a trained machine learning algorithm model that is capable of categorising a milk-producing animal as being of high productivity or low productivity based in input data derived from transcriptomic analysis of a sample obtained from the milk-producing animal, or from one of its parents. Such categorisation facilitates herd-improving management practices as described above.

[0009] Accordingly, the disclosure provides method of predicting the productivity of a milk-producing animal, comprising: (i) quantifying the expression of one or more haemoglobin gene in a sample obtained from the milk-producing animal, or from a parent of the milk-producing animal; and (ii) predicting productivity based on an inverse relationship between quantified expression and productivity.

[0010] The disclosure further provides: a method of using a trained machine learning algorithm model to characterise a milk-producing animal as being of high productivity or low productivity, comprising: (i) receiving input data that comprises data derived from a transcriptomic analysis performed on a sample obtained from the milk-producing animal, or from a parent of the milk-producing animal; and (ii) processing the input data using a trained machine learning algorithm model to generate an output representing information which characterises the milk-producing animal as being of high productivity or low productivity; a method of training a machine learning algorithm model to characterise a milkproducing animal as being of high productivity or low productivity, the method comprising training the machine learning algorithm model using a plurality of training data units, each training data unit comprising (i) input data that comprises data derived from a transcriptomic analysis performed on a sample obtained from a respective training milk-producing animal or from a parent of the respective training milk-producing animal, and (ii) data representing information about the productivity of the respective training milk-producing animal ; a method of selecting a milk-producing animal for breeding, comprising:(a) performing the method of predicting the productivity of a milk-producing animal of the disclosure, and selecting the milk-producing animal or parent for breeding if the method predicts high productivity; or (b) performing the method of using a trained machine learning algorithm model to characterise a milk-producing animal as being of high productivity or low productivity of the disclosure, and selecting the milkproducing animal or parent for breeding if the method characterises the milkproducing animal as being of high productivity.; a computer program comprising instructions that, when executed by a computer system, instruct the computer system to perform the method of using a trained machine learning algorithm model of the disclosure, or the method of training a machine learning algorithm model of the disclosure; and a data processing apparatus comprising a processor configured to perform the method of using a trained machine learning algorithm model of the disclosure, or the method of training a machine learning algorithm model of the disclosure.

[0011] DESCRIPTION OF THE FIGURES

[0012] Figure 1: Schematic outlining the original study design for Example 1. The

[0013] Figure shows number of samples, type of samples, sequencing technology and downstream analysis tools.

[0014] Figure 2: MultiQC plot showing star alignment scores. Each horizontal bar represents one individual in Example 1. Blue (left and middle) shows mapped reads, red (right) shows unmapped reads. All samples (except one) achieved >85% mapping.

[0015] Figure 3: Waterfall plot showing pathways significantly enriched (FDR <0.05) in genes (n=262) which were significantly different (FDR <0.05) between good and poor cows. Blue = pathways linked directly to milk yield, orange = pathways linked to nutrition, red = pathways linked to negative health / stress.

[0016] Figure 4: Violin plots showing gene expression levels (log2-transformed transcripts per million (log2TPM+l). Each violin plot represents the gene expression distribution for a specific gene of interest within the respective sample group. The x-axis denotes the sample group, with "Good Samples" (left) and "Poor Samples" (right) shown. The y-axis represents the gene expression levels in log2TPM+l units. The width of the violin indicates the density of expression values at different levels, with wider sections representing higher density. Box plots are overlaid to show the median expression level and the interquartile range (IQR) encompassing the central 50% of the data.

[0017] Figure 5: Box plots showing gene expression levels (log2-transformed transcripts per million (log2TPM+l). Box plots corresponding to each violin plot in Figure 4 provide a visual summary of the expression distribution, highlighting the median (horizontal line within the box) and the range of values (vertical lines or "whiskers").

[0018] DESCRIPTION OF THE SEQUENCE LISTING

[0019] SEQ ID NO: 1 - DNA sequence of HBA gene (ENSBTAG00000051412).

[0020] SEQ ID NO: 2 - Amino acid sequence of haemoglobin subunit alpha (P01966), encoded by the HBA gene.

[0021] SEQ ID NO: 3 - DNA sequence of HBA1 gene (ENSBTAG00000054844).

[0022] SEQ ID NO: 4 - Amino acid sequence of globin family profile domain-containing protein (A0A452DIQ5), encoded by the HBA1 gene.

[0023] SEQ ID NO: 5 - DNA sequence of HBB gene (ENSBTAG00000037644).

[0024] SEQ ID NO: 6 - Amino acid sequence of haemoglobin subunit beta (P02070), encoded by the HBB gene.

[0025] SEQ ID NO: 7 - DNA sequence of HBM gene (ENSBTAG00000038757).

[0026] SEQ ID NO: 8 - Amino acid sequence of haemoglobin subunit mu (A1A4Q3), encoded by the HBM gene.

[0027] SEQ ID NO: 9 - DNA sequence of HBQ1 gene (ENSBTAG00000016567).

[0028] SEQ ID NO: 10 - Amino acid sequence of globin DI (Al A4Q7), encoded by the HBQ1 gene.

[0029] DETAILED DESCRIPTION It is to be understood that different applications of the disclosed methods, products and uses may be tailored to the specific needs in the art. It is also to be understood that the terminology used herein is for the purpose of describing particular embodiments of the disclosure only, and is not intended to be limiting.

[0030] General definitions

[0031] Unless defined otherwise, technical and scientific terms used herein have the same meaning as commonly understood by a person skilled in the art to which this disclosure belongs.

[0032] As used in this specification and the appended claims, the singular forms “a”, “an”, and “the” include plural referents unless the content clearly dictates otherwise. Thus, for example, reference to “a cow” includes “cows”, reference to “a gene” includes genes”, reference to “a sample” includes “samples”, and the like.

[0033] In general, the term “comprising” is intended to mean including but not limited to. For example, the phrase “input data comprising data derived from a transcriptomic analysis performed on a sample” should be interpreted to mean that the input data contain data derived from a transcriptomic analysis performed on a sample, but that the input data may contain additional components.

[0034] In some aspects of the disclosure, the word “comprising” is replaced with the phrase “consisting of’. The term “consisting of’ is intended to be limiting. For example, the phrase “input data consisting of data derived from a transcriptomic analysis performed on a sample” should be understood to mean that the input data contain data derived from a transcriptomic analysis performed on a sample and no additional components.

[0035] The terms “protein” and “polypeptide” are used interchangeably herein, and are intended to refer to a polymeric chain of amino acids of any length.

[0036] For the purpose of this disclosure, in order to determine the percent identity of two sequences (e.g., two polynucleotide sequences), the sequences are aligned for optimal comparison purposes (e.g., gaps can be introduced in a first sequence for optimal alignment with a second sequence). The nucleotide residues at nucleotide positions are then compared. When a position in the first sequence is occupied by the same nucleotide residue as the corresponding position in the second sequence, then the nucleotides are identical at that position. The percent identity between the two sequences is a function of the number of identical positions shared by the sequences (i.e., % identity = number of identical positions / total number of positions in the reference sequence x 100).

[0037] Typically, the sequence comparison is carried out over the length of the reference sequence. For example, if the user wished to determine whether a given (“test”) sequence has a certain percentage identity to SEQ ID NO: X, SEQ ID NO: X would be the reference sequence. For example, to assess whether a sequence is at least 80% identical to SEQ ID NO: X (an example of a reference sequence), the skilled person would carry out an alignment over the length of SEQ ID NO: X, and identify how many positions in the test sequence were identical to those of SEQ ID NO: X. If at least 80% of the positions are identical, the test sequence is at least 80% identical to SEQ ID NO: X. If the sequence is shorter than SEQ ID NO: X, the gaps or missing positions should be considered to be non-identical positions.

[0038] The skilled person is aware of different computer programmes that are available to determine the homology or identity between two sequences. For instance, a comparison of sequences and determination of percent identity between two sequences can be accomplished using a mathematical algorithm.

[0039] Method of predicting productivity

[0040] The disclosure provides method of predicting the productivity of a milk-producing animal, comprising: (i) quantifying the expression of one or more haemoglobin gene in a sample obtained from the milk-producing animal, or from a parent of the milk-producing animal; and (ii) predicting productivity based on an inverse relationship between quantified expression and productivity. As explained above, the predicted productivity can be used to determine how the milk-producing animal or parent is managed. For instance, a high productivity milk-producing animal may be retained in the herd, or a low productivity milk-producing animal may be removed from the herd, with a view of improving productivity at the herd level. Productivity of the herd may also be improved by selecting high productivity animals, or a parent giving rise to high productivity offspring, for use in breeding new herd members. Improvements in productivity have important implications for sustainable farming and the economic viability of the herd.

[0041] Productivity

[0042] In the context of the present application, the productivity of a milk-producing animal is a measure of one or more characteristics of milk production by the milkproducing animal. Characteristics indicative of high productivity are known in the art and include, for example, milk yield, milk fat percentage, milk protein percentage, and milking speed.

[0043] Productivity as predicted by the method of the disclosure may be a measure of any one or more of these. For instance, productivity may be a measure of (i) milk yield, (ii) milk fat percentage, (iii) milk protein percentage, or (iv) milking speed. Productivity may be a measure of: (i) and (ii); (i) and (iii); (i) and (iv); (ii) and (iii); (ii) and (iv); (iii) and (iv); (i), (ii) and (iii); (i), (ii) and (iv); (ii), (iii) and (iv); or (i), (ii), (iii) and (iv). The method of the disclosure may predict (i); (ii); (iii); (iv); (i) and (ii); (i) and (iii); (i) and (iv); (ii) and (iii); (ii) and (iv); (iii) and (iv); (i), (ii) and (iii); (i), (ii) and (iv); (ii), (iii) and (iv); or (i), (ii), (iii) and (iv). The method of the disclosure may further predict one or more additional characteristics indicative of high productivity.

[0044] In any case, milk yield refers to the quantity of milk that may be obtained from an individual milk-producing animal in a period of time. Quantity may, for example, be measured in kilograms or litres. The period of time may, for example, be a day or a year. Thus, milk yield may be expressed in terms of kg / day, litres / day, kg / year, or litres / year, for example. As a dairy farmer’s income often depends on the quantity of milk produced, it is desirable to maximise milk yield. Improving milk yield may also have an effect on sustainability. Therefore, a higher milk yield is desirable. Milk yield can vary significantly between individuals and farms. Milk yield may, for instance, differ with breed. For example, a Holstein-Friesian cow will typically have a higher milk yield than a Jersey or Guernsey cow. Milk yield may, for instance differ with husbandry practices and / or farm infrastructure. For instance, milk-producing animals milked more than twice a day may produce more milk than milk-producing animals milked twice-daily. Robotic milking parlours may increase the frequency of milking and, therefore the yield from individual milk-producing animals.

[0045] In general terms, though, a milk yield of 35kg / day or more, for example, may be considered to be a high milk yield for cattle. For instance, a milk yield of 40 kg / day or more, 45 kg / day or more, 50 kg / day or more, 55 kg / day or more, 60 kg / day or more, 70 kg / day or more, 75 kg / day or more, 80 kg / day or more, 85 kg / day or more, 90kg / day or more, 95 kg / day or more or 100 kg / day or more may be considered a high milk yield. A yield of 35 to 100 kg / day, such as 40 to 95 kg / day, 45 to 90 kg / day, 50 to 85 kg / day, 55 to 80 kg / day, 60 to 75 kg / day, or 65 to 70 kg / day, may, for example, be considered to be a high milk yield. In general terms, though, a milk yield of less than 35kg / day, for example, may be considered to be a low milk yield for cattle. For instance, a milk yield of less than 30 kg / day, less than 25 kg / day, less than 20 kg / day, less than 15 kg / day, less than 10 kg / day, or less than 5 kg / day may be considered a low milk yield. A yield of 0 to 35 kg / day, such as 5 to 30 kg / day, 10 to 25 kg / day, or 15 to 20 kg / day, may, for example, be considered to be a low milk yield. Exemplary high and low milk yields are also well- known in the art for goats and sheep.

[0046] Milk yield may, alternatively, be assessed on a relative basis, to account for differences between individuals, herds and farms. For instance, a high milk yield may be taken to mean a milk yield that is above the average milk yield for the herd to which an individual animal belongs. A low milk yield may, for example, be taken to mean a milk yield that is below the average milk yield for the herd to which an individual animal belongs. The average may be the mean, median or mode. Typically, the average is the mean. Relative assessment of milk yield may be applied to any bovid.

[0047] Milk fat, otherwise known as butterfat, is the fatty portion of milk. The price that a farmer receives for milk sold from their farm is often proportionate to the percentage milk fat it contains. Furthermore, a higher milk fat content increases the nutritional value of milk. Therefore, a higher percentage milk fat is desirable. Milk fat can vary significantly between individuals and farms. Milk fat may, for instance, differ with breed. For example, a Jersey or Guernsey cow will typically have a higher milk fat than a Holstein- Friesian cow. Milk fat may, for instance differ with husbandry practices and / or farm infrastructure. Feeding regime and / or nutrition may, for example, affect milk fat. In general terms, though, a milk fat of 4.2 % or more, for example, may be considered to be a high milk fat for cattle. For instance, a milk fat of 4.1 % or more, 4.3 % or more, 4.4 % or more, 4.5 % or more, 4.6 % or more, 4.7 % or more, 4.8 % or more, 4.9 % or more, 5.0 % or more, 5.1 % or more, 5.2 % or more, 5.3 % or more, 5.4 % or more, 5.5 % or more, 5.6 % or more 5.7 % or more 5.8 % or more, 5.9 % or more or 6 % or more may be considered to be a high milk fat. A milk fat of 4.2 to 6 %, such as 4.3 to 5.9 %, 4.4 to 5.8 %, 4.5 to 5.7 %, 4.6 to 5.6 %, 4.7 to 5.5 %, 4.8 to 5.4 %, 4.9 to 5.3 %, or 5.0 to 5.2, may, for example, be considered to be a high milk fat. In general terms, though, a milk fat of less than 3.2 %, for example, may be considered to be a low milk fat for cattle. For instance, a milk fat of less than 3.1 %, less than 3.0 %, less than 2.9 %, less than 2.8 %, less than 2.7 %, less than 2.6 %, less than 2.5 %, less than 2.4 %, less than 2.3 %, less than 2.2 %, less than 2.1 %, or less than 2.0 % may be considered to be a low milk fat. A fat of 2.0 to 3.2 %, such as 2.1 to 3.1 %, 2.2 to 3.0 %, 2.3 to 2.9 %, 2.4 to 2.8 %, or 2.5 to 2.7 % may, for example, be considered to be a low milk fat. Exemplary high and low milk fat percentages are also well-known in the art for goats and sheep.

[0048] Milk fat may, alternately, be assessed on a relative basis, to account for differences between individuals, herds and farms. For instance, a high milk fat may be taken to mean a milk fat that is above the average milk fat for the herd to which an individual animal belongs. A low milk fat may, for example, be taken to mean a milk fat that is below the average milk fat for the herd to which an individual animal belongs. The average may be the mean, median or mode. Typically, the average is the mean. Relative assessment of milk fat may be applied to any milk-producing animal, such as any bovid.

[0049] Milk protein is the total protein content of milk. The proteins found in milk typically include casein and whey protein. Whey protein may comprise a-lactalbumin, P- lactoglobulin, serum albumin, and / or immunoglobulin, for example. The price that a farmer receives for milk sold from their farm is often proportionate to the percentage milk protein it contains. Furthermore, a higher milk protein content increases the nutritional value of milk. Therefore, a higher percentage milk protein is desirable. Milk protein can vary significantly between individuals and farms. Milk protein may, for instance, differ with breed. For example, a Jersey or Guernsey cow will typically have a higher milk protein than a Holstein-Friesian cow. Milk protein may, for instance differ with husbandry practices and / or farm infrastructure. Feeding regime and / or nutrition may, for example, affect milk protein.

[0050] In general terms, though, a total milk protein of 3.5 % or more, for example, may be considered to be a high milk protein for cattle. For instance, a milk protein of 3.6 % or more, 3.7 % or more, 3.8 % or more, 3.9 % or more, 4.0 % or more, 4.1 % or more, 4.2 % or more, 4.3 % or more, 4.3 % or more, 4.4 % or more, 4.6 % or more 4.7 % or more 4.8 % or more, 4.9 % or more or 5 % or more may be considered a high milk protein. A milk protein of 3.5 to 5 %, such as 3.6 to 4.9 %, 3.7 to 4.8 %, 3.8 to 4.7 %, 3.9 to 4.6 %, 4.0 to 4.4 %, or 4.1 to 4.3 % may, for example, be considered to be a high milk protein. In general terms, though, a milk protein of less than 2.8 %, for example, may be considered to be a low milk protein for cattle. For instance, a milk protein of less than 2.7 %, less than 2.6 %, less than 2.5 %, less than 2.4 %, less than 2.3 %, less than 2.2 %, less than 2.1 %, less than 2.0 %, less than 1.9 %, less than 1.8 % or less than 1.7 % may be considered a low milk protein. A protein of 1.7 to 2.8 %, such as 1.8 to 2.7 %, 1.9 to 2.6 %, 2.0 to 2.5 %, 2.1 to 2.4 %, or 2.2 to 2.3 %, may, for example, be considered to be a low milk protein. Exemplary high and low milk protein percentages are also well-known in the art for goats and sheep.

[0051] Milk protein may, alternately, be assessed on a relative basis, to account for differences between individuals, herds and farms. For instance, a high milk protein may be taken to mean a milk protein that is above the average milk protein for the herd to which an individual animal belongs. A low milk protein may, for example, be taken to mean a milk protein that is below the average milk protein for the herd to which an individual animal belongs. The average may be the mean, median or mode. Typically, the average is the mean. Relative assessment of milk protein may be applied to any milk-producing animal, such as any bovid.

[0052] Milking speed is the rate at which milk may be obtained from a milk-producing animal. To obtain milk from a milk-producing animal, the milk-producing animal is milked one or more times (typically two or more times) per day. Typically, a milking machine is used. Milking speed may be determined by measuring the quantity of milk obtained over a unit of time during milking. The ability to make such determination is inbuilt to many milking machines. Many milking machines are also able to determine milk yield, milk fat and / or milk protein. A milk-producing animal that milks slowly may disrupt the flow of milk-producing animals through a milking parlour. A milk-producing animal that milks quickly causes less disruption, and may require less labour to milk. Therefore, a faster milking speed is desirable. Milking speed can vary significantly between individuals and farms. Milking speed may, for instance, differ with breed. For example, a Holstein-Friesian cow will typically have a higher milking speed than a Jersey or Guernsey cow. Milking speed may, for instance differ with husbandry practices and / or farm infrastructure. The type of milking equipment used may, for example, affect milking speed.

[0053] In general terms, though, a milking speed of 3.0 1 / min or more, for example, may be considered to be a high milking speed for cattle. For instance, a milking speed of 3.1 1 / min or more, 3.2 1 / min or more, 3.3 1 / min or more, 3.4 1 / min or more, 3.5 1 / min or more,

[0054] 3.6 1 / min or more, 3.7 1 / min or more, 3.8 1 / min or more, 3.9 1 / min or more, 4.0 1 / min or more, 4.1 1 / min or more 4.2 1 / min or more 4.3 1 / min or more, 4.4 1 / min or more or 4.5 1 / min or more may be considered a high milking speed. A milking speed of 3.0 to 4.5 1 / min, such as 3.1 to 4.4 1 / min, 3.2 to 4.3 1 / min, 3.3 to 4.2 1 / min, 3.4 to 4.1 1 / min, 3.5 to 4.0 1 / min,

[0055] 3.6 to 3.9 1 / min, or 3.7 to 3.8 1 / min may, for example, be considered to be a high milking speed. In general terms, though, a milking speed of less than 2.4 1 / min, for example, may be considered to be a low milking speed for cattle. For instance, a milking speed of less than 2.3 1 / min, less than 2.2 1 / min, less than 2.2 1 / min, less than 2.1 1 / min, less than 2.0 1 / min, less than 1.9 1 / min, less than 1.8 1 / min, less than 1.7 1 / min, less than 1.6 1 / min, less than 1.5 1 / min or less than 1.4 1 / min may be considered a low milking speed. A protein of

[0056] 1.7 to 2.8 1 / min, such as 1.8 to 2.7 1 / min, 1.9 to 2.6 1 / min, 2.0 to 2.5 1 / min, 2.1 to 2.4 1 / min, or 2.2 to 2.3 1 / min, may, for example, be considered to be a low milking speed. Exemplary high and low milking speeds are also well-known in the art for goats and sheep.

[0057] Milking speed may, alternately, be assessed on a relative basis, to account for differences between individuals, herds and farms. For instance, a high milking speed may be taken to mean a milking speed that is above the average milking speed for the herd to which an individual animal belongs. A low milking speed may, for example, be taken to mean a milking speed that is below the average milking speed for the herd to which an individual animal belongs. The average may be the mean, median or mode. Typically, the average is the mean. Relative assessment of milk protein may be applied to any milkproducing animal, such as any bovid.

[0058] Accordingly, high productivity may be defined as (i) high milk yield, (ii) high milk fat percentage, (iii) high milk protein percentage, and / or (iv) high milking speed. High productivity may, for instance, be: (i); (ii); (iii); (iv); (i) and (ii); (i) and (iii); (i) and (iv);

[0059] (ii) and (iii); (ii) and (iv); (iii) and (iv); (i), (ii) and (iii); (i), (ii) and (iv); (ii), (iii) and (iv); or (i), (ii), (iii) and (iv). The method of the disclosure may predict (i); (ii); (iii); (iv); (i) and (ii); (i) and (iii); (i) and (iv); (ii) and (iii); (ii) and (iv); (iii) and (iv); (i), (ii) and (iii);

[0060] (i), (ii) and (iv); (ii), (iii) and (iv); or (i), (ii), (iii) and (iv). Preferably, high productivity is defined as (i) high milk yield, and optionally one or more of (ii) high milk fat percentage,

[0061] (iii) high milk protein percentage, and (iv) high milking speed. For instance, high productivity may, for example, be defined as: (i); (i) and (ii); (i) and (iii); (i) and (iv);; (i),

[0062] (ii) and (iii); (i), (ii) and (iv); or (i), (ii), (iii) and (iv).

[0063] Low productivity may be defined as (I) low milk yield, (II) low milk fat percentage, (III) low milk protein percentage, and / or (IV) low milking speed. Low productivity may, for instance, be: (I); (II); (III); (IV); (I) and (II); (I) and (III); (I) and (IV); (II) and (III); (II) and (IV); (III) and (IV); (I), (II) and (III); (I), (II) and (IV); (II), (III) and (IV); or (I), (II), (III) and (IV). The method of the disclosure may predict (I); (II); (III); (IV); (I) and (II);

[0064] (I) and (III); (I) and (IV); (II) and (III); (II) and (IV); (III) and (IV); (I), (II) and (III); (I),

[0065] (II) and (IV); (II), (III) and (IV); or (I), (II), (III) and (IV). Preferably, low productivity is defined as (I) low milk yield, and optionally one or more of (II) low milk fat percentage,

[0066] (III) low milk protein percentage, and (IV) low milking speed. For instance, low productivity may, for example, be defined as: (I); (I) and (II); (I) and (III); (I) and (IV); (I), (II) and (III); (I), (II) and (IV); or (I), (II), (III) and (IV).

[0067] Milk-producing animal

[0068] The method of the disclosure predicts the productivity of a milk-producing animal. In the context of the disclosure, a milk-producing animal is an animal that belongs to a species that is capable of producing milk. Typically, the species is used for commercial milk production. Accordingly, the milk-producing animal may be an animal that produces milk, or is capable of producing milk. By way of example, a juvenile milk-producing animal that has not yet produced offspring may not yet produce milk, but is capable of doing so.

[0069] Preferably, the milk-producing animal is non-human animal.

[0070] The milk-producing animal may, for example, be a bovid. That is, the milk producing animal may be a member of the Bovidae family. The bovid may, for example, be a bovine. In other words, the bovid may be a member of the subfamily Bovinae. Bovinae include, for example, cattle, bison and buffalo. The bovid may, for example, be a caprine. In other words, the bovid may be a member of the subfamily Caprinae. Caprinae include goats and sheep.

[0071] The milk-producing animal may, for example, be a Bos taurus. Bos taurus are commonly known as cattle. The Bos Taurus may, for example, be a cow or heifer. In the context of the present application, the terms cow and heifer both refer to females of the Bos taurus species. A cow is a mature female Bos taurus that has given birth. A heifer is a female Bos taurus that has yet to give birth. However, in the prior art, the term cow may be used to refer generically to any Bos taurus individual.

[0072] Many different breeds of cattle exist, including breeds that are particularly suited to milk production (dairy breeds) and breeds that are particularly suited to meat production (beef breeds). In a preferred aspect of the disclosure, the Bos taurus, or belongs to, a dairy breed. However, the Bos taurus may alternatively be, or belong to, a beef breed. Milk production is an important productivity trait for beef cattle as well as dairy cattle, because it is relevant to the mothering ability of beef cows and thus the growth of beef calves.

[0073] Numerous dairy breeds are well-known in the art. Common dairy breeds include Holstein Friesian (or Holstein), Jersey, Guernsey, Brown Swiss, Ayrshire, and Shorthorn. Other dairy breeds include Abondance, Alderney, American Milking Devon, Australian Braford, Australian Friesian Sahiwal, Australian Lowline, Australian Milking Zebu, Belgian Red, Bianca Modenese, Burlina, Busa, Canadienne, Danish Jersey, Danish Red, Dexter, Estonian Red, Fleckvieh (or Simmental), French Simmental, German Black Pied Dairy, Girolando, Harzer Rotvieh, Hays Converter, Herens, Illawarra Shorthorn, Irish Moiled, Jamaica Hope, Lakenvelder (or Dutch Belted), Meuse-Rhine-Issel, Montbeliarde, Normande, Norwegian Red, Old Gloucester, Randall Lineback, Pie Rouge des Plaines, Pinzgauer, Red Chittagong, Red Poll, Red Sindhi, Sahiwal, Sussex, Tipo Carora, Tyrol Grey, and Vorderwald. The Bos taurus may be, or belong to, any of these breeds. The Bos taurus may be a cross between any of these breeds.

[0074] Numerous beef breeds are well-known in the art. Beef breeds include Aberdeen Angus, Hereford, Lincoln Red, Dexter, Belted Galloway, Reg Angus, Red Poll, Highland, Shorthorn, Charolais, Chianina, Fleckvieh (or Simmental), Gelbvieh, Braunvieh, Limousin, Blonde d’ Aquitaine, Piedmontese, Romagnola, Maine Anjou, Murray Grey, Tarentaise, Belgian Blue, Wagyu, Salers, Texas Longhorn, Watusi. The Bos taurus may be, or belong to, any of these breeds. The Bos taurus may be a cross between any of these breeds.

[0075] Some cattle breeds are particularly suited to both milk production and meat production, i.e. are dual purpose. Thus, in the context of the present application, they may be considered a dairy breed or a beef breed. Dual purpose breeds include Dexter, Fleckvieh (or Simmental), and Shorthorn.

[0076] The Bos taurus may be a cross between a dairy breed and a beef breed.

[0077] A Bos taurus that is, or belongs to, a dairy breed typically has parents that are, or belong to, a dairy breed. For example, one or both parents of a Bos taurus that is, or belongs to, a dairy breed may be, or belong to, the same dairy breed. Accordingly, one or both parents of the Bos taurus may be, or belong to, a dairy breed.

[0078] A Bos taurus that is, or belongs to, a beef breed typically has parents that are, or belong to, a beef breed. For example, one or both parents of a Bos taurus that is, or belongs to, a beef breed may be, or belong to, the same beef breed. Accordingly, one or both parents of the cow or heifer may be, or belong to, a beef breed.

[0079] A Bos taurus that is a cross between a dairy breed and a beef breed may, for example, be the offspring of a parent that is, belongs to, a dairy breed and a parent that is, belongs to, a beef breed. Accordingly, one parent of the cow or heifer may be, or belong to, a dairy breed and one parent of the cow or heifer may be, or belong to, a beef breed.

[0080] The milk-producing animal may, for example, be a Capra hircus. Capra hircus are commonly known as goats. The Capra hircus may, for example, be a doe or a doe kid. In the context of the present application, the terms doe and doe kid both refer to females of the Capra hircus species. A doe is a mature female Capra hircus. A doe kid is an immature female Capra hircus. However, in the prior art, the term doe may be used to refer generically to any Capra hircus female.

[0081] Many different breeds of goat exist, including breeds that are particularly suited to milk production (dairy breeds) and breeds that are particularly suited to meat production (meat breeds). In a preferred aspect of the disclosure, the Capra hircus or belongs to, a dairy breed. However, the Capra hircus may alternatively be, or belong to, a meat breed. Milk production is an important productivity trait for meat breeds as well as dairy breeds, because it is relevant to the mothering ability of meat goats and thus the growth of kids.

[0082] Numerous dairy breeds are well-known in the art. Common dairy breeds include Alpine, Saanen, LaMancha, Jamunapari, Toggenburg, Nubian, Oberhasli, Guernsey, Sable and Nigerian Dwarf. The Capra hircus may be, or belong to, any of these breeds. The Capra hircus may be a cross between any of these breeds.

[0083] Numerous meat breeds are well-known in the art. Meat breeds include Black Bengal, Boar, Kalahari, Kiko, Fainting, Nubian, Rangeland, Sirohi, Spanish and Verata. The Capra hircus may be, or belong to, any of these breeds. The Capra hircus may be a cross between any of these breeds.

[0084] Some goat breeds are particularly suited to both milk production and meat production, i.e. are dual purpose. Thus, in the context of the present application, they may be considered a dairy breed or a meat breed. Dual purpose breeds include the Nubiamn.

[0085] The Capra hircus may be a cross between a dairy breed and a meat breed.

[0086] A Capra hircus that is, or belongs to, a dairy breed typically has parents that are, or belong to, a dairy breed. For example, one or both parents of a Capra hircus that is, or belongs to, a dairy breed may be, or belong to, the same dairy breed. Accordingly, one or both parents of the Capra hircus may be, or belong to, a dairy breed.

[0087] A Capra hircus that is, or belongs to, a meat breed typically has parents that are, or belong to, a meat breed. For example, one or both parents of a Capra hircus that is, or belongs to, a meat breed may be, or belong to, the same meat breed. Accordingly, one or both parents of the Capra hircus may be, or belong to, a meat breed.

[0088] A Capra hircus that is a cross between a dairy breed and a meat breed may, for example, be the offspring of a parent that is, belongs to, a dairy breed and a parent that is, belongs to, a meat breed. Accordingly, one parent of the Capra hircus may be, or belong to, a dairy breed and one parent of the Capra hircus may be, or belong to, a meat breed.

[0089] The milk-producing animal may, for example, be a Ovis aries. Ovis aries are commonly known as sheep. The Ovis aries may, for example, be a ewe, gimmer or ewe lamb. In the context of the present application, the terms ewe, gimmer and ewe lamb each refer to females of the Ovis aries species. A ewe is a mature female Ovis aries that has given birth. A gimmer is a female Ovis aries that has yet to give birth, that is typically in its second year of life. A gimmer may also be known as a theave. A ewe lamb is an immature female Ovis aries, typically younger than a gimmer.

[0090] Many different breeds of sheep exist, including breeds that are particularly suited to milk production (dairy breeds) and breeds that are particularly suited to meat production (meat breeds). In a preferred aspect of the disclosure, the Ovis aries or belongs to, a dairy breed. However, the Ovis aries may alternatively be, or belong to, a meat breed. Milk production is an important productivity trait for meat breeds as well as dairy breeds, because it is relevant to the mothering ability of meat sheep and thus the growth of lambs.

[0091] Numerous dairy breeds are well-known in the art. Common dairy breeds include East Friesian, Sarda, Lacaune, British Milk Sheep, Chios, Awassi, Assaf, and Zwartbles. The Ovis aries may be, or belong to, any of these breeds. The Ovis aries may be a cross between any of these breeds.

[0092] Numerous meat breeds are well-known in the art. Meat breeds include Cheviot, Dorset, Hampshire, Montadale, North County Cheviot, Border Leicester, Dorper, Wiltshire Horn, Oxford, Shropshire, Southdown, Suffolk, Texel, Blue Texel, Tunis and Zwartbles. The Ovis aries may be, or belong to, any of these breeds. The Ovis aries may be a cross between any of these breeds.

[0093] Some sheep breeds are particularly suited to both milk production and meat production, i.e. are dual purpose. Thus, in the context of the present application, they may be considered a dairy breed or a meat breed. Dual purpose breeds include the Zwartbles.

[0094] The Ovis aries may be a cross between a dairy breed and a meat breed.

[0095] An Ovis aries that is, or belongs to, a dairy breed typically has parents that are, or belong to, a dairy breed. For example, one or both parents of an Ovis aries that is, or belongs to, a dairy breed may be, or belong to, the same dairy breed. Accordingly, one or both parents of the Ovis aries may be, or belong to, a dairy breed.

[0096] An Ovis aries that is, or belongs to, a meat breed typically has parents that are, or belong to, a meat breed. For example, one or both parents of a Ovis aries that is, or belongs to, a meat breed may be, or belong to, the same meat breed. Accordingly, one or both parents of the Ovis aries may be, or belong to, a meat breed.

[0097] An Ovis aries that is a cross between a dairy breed and a meat breed may, for example, be the offspring of a parent that is, belongs to, a dairy breed and a parent that is, belongs to, a meat breed. Accordingly, one parent of the Ovis aries may be, or belong to, a dairy breed and one parent of the Ovis aries may be, or belong to, a meat breed.

[0098] Sample

[0099] The method comprises quantifying the expression of one or more haemoglobin gene in a sample obtained from the milk-producing animal, or from a parent of the milkproducing animal.

[0100] Use of a sample obtained from the milk-producing animal allows for prediction of productivity of that milk-producing animal, which already exists.

[0101] Use of a sample from a parent of the milk-producing animal allows for prediction of productivity of the female offspring, or potential female offspring, of that parent. The parent may, for instance, be a prospective parent (e.g. sire or dam), and the milk-producing animal may be its potential progeny. In this case, predicting the productivity of the milkproducing animal may assist in deciding whether or not to use the parent for breeding.

[0102] The sample may be any sample that comprises mRNA or protein expressed by the milk-producing animal or parent. The sample may, for example, comprise cells belonging to the milk-producing animal or parent.

[0103] The sample may, for example, be a blood sample. A blood sample comprises cells, such as leukocytes, that express mRNA and protein.

[0104] The sample may, for example, be a milk sample. A milk sample comprises cells, such as leukocytes and / or epithelial cells, that express mRNA and protein.

[0105] The sample may, for example, be a hair sample. A hair sample comprises cells, that express mRNA and protein, such as cells from the hair root or hair follicle. The sample may, for example, be a saliva sample. A saliva sample comprises cells, such as leukocytes and / or epithelial cells, that express mRNA and protein.

[0106] The sample may, for example, be a tissue sample. For instance, the tissue sample may be a cheek swab sample. A cheek swab sample comprises cells, such as epithelial cells, that express mRNA and protein.

[0107] When the sample is obtained from a parent of the milk-producing animal, and the parent is a male, the sample may be a semen sample, for example. In addition to spermatozoa, a semen sample comprises leukocytes and epithelial cells.

[0108] Quantifying expression of a gene

[0109] Expression of one or more haemoglobin gene in the sample may be quantified by any suitable means. Means for quantifying gene expression are well-known in the art.

[0110] In a preferred aspect of the invention, expression of one or more haemoglobin gene is quantified by measuring expression of the mRNA encoded by the haemoglobin gene. Means for measuring mRNA expression are known in the art. mRNA expression may, for example, be measured using RNA-Seq, a microarray, qPCR, Northern blot, next generation sequencing (NGS), or digital PCR.

[0111] Expression of one or more haemoglobin gene may alternatively be quantified by measuring expression of the protein encoded by the haemoglobin gene. Means for measuring protein expression are known in the art. Protein expression may, for example, be measured by Western blot, immunocytochemistry, immunohistochemistry, flow cytometry, or mass spectrometry.

[0112] Any of these techniques are within the routine skill set of the person skilled in the art.

[0113] One or more haemoglobin gene

[0114] The method comprises quantifying the expression of one or more haemoglobin gene in a sample obtained from the milk-producing animal, or from a parent of the milkproducing animal. The method may comprise quantifying the expression of two or more, three or more, four or more, or five or more haemoglobin genes in a sample obtained from the milk-producing animal, or from a parent of the milk-producing animal. Haemoglobin is the iron-containing oxygen-transport protein present in red blood cells of almost all vertebrates. Haemoglobin comprises multiple protein subunits, known as globin molecules. Constituent subunits may comprise a ring-like heme group containing an iron atom. Oxygen binds reversibly to these iron atoms and is transported through blood.

[0115] In cattle, there are multiple genes encoding globin molecules. These comprise HBA, HBA1, HBB, HBM and HBQ1. HBA has the Ensembl Gene ID ENSBTAG00000051412, and the sequence of SEQ ID NO: 1. HBA encodes haemoglobin subunit alpha (SEQ ID NO: 2), having the UniProt reference P01966. HBA1 has the Ensembl Gene ID ENSBTAG00000054844, and the sequence of SEQ ID NO: 3. HBA1 encodes globin family profile domain-containing protein (SEQ ID NO: 4), having the UniProt reference A0A452DIQ5. HBB has the Ensembl Gene ID ENSBTAG00000037644, and the sequence of SEQ ID NO: 5. HBB encodes haemoglobin subunit beta (SEQ ID NO: 6), having the UniProt reference P02070. HBM has the Ensembl Gene ID ENSBTAG00000038757, and the sequence of SEQ ID NO: 7. HBM encodes haemoglobin subunit mu (SEQ ID NO: 8), having the UniProt reference A1A4Q3. HBQ1 has the Ensembl Gene ID ENSBTAG00000016567, and the sequence of SEQ ID NO: 9. HBQ1 encodes globin DI (SEQ ID NO: 10), having the UniProt reference Al A4Q7. Similar genes may exist in other bovids, such as goats and sheep.

[0116] The method may comprise quantifying the expression of any one or more of (a) HBA or an equivalent gene in a non-cattie bovid, (b) HBA1 or an equivalent gene in a noncattle bovid, (c) HBB or an equivalent gene in a non-cattie bovid, (d) HBM or an equivalent gene in a non-cattie bovid; and (e) HBQ1 or an equivalent gene in a non-cattie bovid. For instance, the method may comprise quantifying the expression of: (a); (b); (c); (d); (e); (a) and (b); (a) and (c); (a) and (d); (a) and (e); (b) and (c); (b) and (d); (b) and (e);

[0117] (c) and(d); (c) and (e); (d) and (e); (a), (b) and (c); (a), (b) and (d); (a), (b) and (e); (a), (c) and (d); (a), (c) and (e); (a), (d) and (e); (b), (c) and (d); (b), (c) and (e); (b), (d) and (e); (c),

[0118] (d) and (e); (a), (b), (c) and (d); (a), (b), (c) and (e); (a), (b), (d) and (e); (a), (c), (d) and (e); (b), (c), (d) and (e); (a), (b), (c), (d) and (e). In one aspect of the disclosure, the method comprises quantifying the expression of (a), (b), (c), (d) and (e). The method may comprise quantifying the expression of any one or more of (A) a gene having at least 90% identity to HBA (SEQ ID NO: 1); (B) a gene having at least 90% identity to HBA1 (SEQ ID NO: 3); (C) a gene having at least 90% identity to HBB (SEQ ID NO: 5); (D) a gene having at least 90% identity to HBM (SEQ ID NO: 7); and (E) a gene having at least 90% identity to HBQ1 (SEQ ID NO: 9). For instance, the method may comprise quantifying the expression of: (For instance, the method may comprise quantifying the expression of: (A); (B); (C); (D); (E); (A) and (B); (A) and (C); (A) and

[0119] (D); (A) and (E); (B) and (C); (B) and (D); (B) and (E); (C) and (D); (C) and (E); (D) and

[0120] (E); (A), (B) and (C); (A), (B) and (D); (A), (B) and (E); (A), (C) and (D); (A), (C) and (E); (A), (D) and (E); (B), (C) and (D); (B), (C) and (E); (B), (D) and (E); (C), (D) and (E);

[0121] (A), (B), (C) and (D); (A), (B), (C) and (E); (A), (B), (D) and (E); (A), (C), (D) and (E);

[0122] (B), (C), (D) and (E); (A), (B), (C), (D) and (E). In one aspect of the disclosure, the method comprises quantifying the expression of (A), (B), (C), (D) and (E). In (A), the gene may, for example, have at least 95%, at least 96%, at least 97%, at least 98%, or at least 99% identity to HBA (SEQ ID NO: 1). In (B), the gene may, for example, have at least 95%, at least 96%, at least 97%, at least 98%, or at least 99% identity to HBA1 (SEQ ID NO: 3). In (C), the gene may, for example, have at least 95%, at least 96%, at least 97%, at least 98%, or at least 99% identity to HBB (SEQ ID NO: 5). In (D), the gene may, for example, have at least 95%, at least 96%, at least 97%, at least 98%, or at least 99% identity to HBM (SEQ ID NO: 7). In (E), the gene may, for example, have at least 95%, at least 96%, at least 97%, at least 98%, or at least 99% identity to HBQ1 (SEQ ID NO: 9).

[0123] As explained above, expression may be quantified by measuring expression of the mRNA or protein encoded by the relevant gene. In the case that expression of the protein is measured, the method may comprise quantifying the expression of any one or more of

[0124] (A) a protein having at least 90% identity to the protein (SEQ ID NO: 2) encoded by HBA;

[0125] (B) a protein having at least 90% identity to the protein (SEQ ID NO: 4) encoded by HBA1; (C) a protein having at least 90% identity to the protein (SEQ ID NO: 6)encoded by HBB; (D) a protein having at least 90% identity to the protein (SEQ ID NO: 8)encoded by HBM; and (E) a protein having at least 90% identity to the protein (SEQ ID NO: 10) encoded by HBQ1. For instance, the method may comprise quantifying the expression of: (I); (II); (III); (IV); (V); (I) and (II); (I) and (III); (I) and (IV); (I) and (V); (II) and (III); (II) and (IV); (II) and (V); (III) and(IV); (III) and (V); (IV) and (V); (I), (II) and (III); (I), (II) and (IV); (I), (II) and (V); (I), (III) and (IV); (I), (III) and (V); (I), (IV) and (V); (II), (III) and (IV); (II), (III) and (V); (II), (IV) and (V); (III), (IV) and (V); (I), (II), (III) and (IV); (I), (II), (III) and (V); (I), (II), (IV) and (V); (I), (III), (IV) and (V); (II), (III), (IV) and (V); (I), (II), (III), (IV) and (V) . In one aspect of the disclosure, the method comprises quantifying the expression of (I), (II), (III), (IV) and (V). In (I), the protein may, for example, have at least 95%, at least 96%, at least 97%, at least 98%, at least 99%, or 100% identity to the protein (SEQ ID NO: 2) encoded by HBA. In (II), the protein may, for example, have at least 95%, at least 96%, at least 97%, at least 98%, at least 99%, or 100% identity to the protein (SEQ ID NO:4) encoded by HBA1. In (III), the protein may, for example, have at least 95%, at least 96%, at least 97%, at least 98%, at least 99%, or 100% identity to the protein (SEQ ID NO: 6) encoded by HBB. In (IV), the protein may, for example, have at least 95%, at least 96%, at least 97%, at least 98%, at least 99%, or 100% identity to the protein (SEQ ID NO: 8) encoded by HBM. In (V), the protein may, for example, have at least 95%, at least 96%, at least 97%, at least 98%, at least 99%, or 100% identity to the protein (SEQ ID NO: 10) encoded by HBQ1.

[0126] Predicting productivity

[0127] Whereas some have sought to correlate levels of glycosylated haemoglobin with milk and lactose production traits in dairy cattle, the present inventors have, for the first time, demonstrated a surprising association between expression of one or more haemoglobin gene and the productivity of a milk-producing animal. In particular, there is an inverse relationship between the level of expression of the one or more haemoglobin gene in a sample obtained from the milk-producing animal (or a parent of the milkproducing animal) and productivity. That is, as expression of the one or more haemoglobin gene decreases, productivity increases. As expression of the one or more haemoglobin gene increases, productivity decreases. Therefore, a milk-producing animal with lower expression of one or more haemoglobin gene may perform well in terms of milk yield, milk fat percentage, milk protein percentage, milking speed or other indicators of productivity. A milk-producing animal with a parent having lower expression of one or more haemoglobin gene may perform well in terms of milk yield, milk fat percentage, milk protein percentage, milking speed or other indicators of productivity.

[0128] Accordingly, the method of the disclosure comprises predicting productivity based on an inverse relationship between quantified expression of the one or more haemoglobin gene and productivity. In other words, once expression of the one or more haemoglobin gene is quantified, productivity is predicted based on the level of expression. A lower level of expression predicts higher productivity, and a higher level of expression predicts lower productivity. Thus, the output of the method may be an expression level, which is itself an indicator of productivity. This may apply whether mRNA or protein expression is quantified. For example, when mRNA expression is quantified using RNA-Seq, the output of the method may be a count for each of the one or more haemoglobin genes (i.e. a number of reads mapped onto each of the one or more haemoglobin genes). Lower counts predict higher productivity, and higher counts predict lower productivity. In another example, when protein expression is quantified using a Western blot, the output of the method may be the concentration of each of the one or more haemoglobin genes relative to the concentration of a housekeeping protein used as a loading control. Housekeeping proteins are known in the art and include P-actin, GAPDH and HPRT1, for example. Concentration may be measured by measuring the signal intensity of protein bands in the blot. Lower concentrations of haemoglobin proteins predict higher productivity, and higher concentrations of haemoglobin proteins predict lower productivity.

[0129] The method of the disclosure may therefore produce an expression level, which functions as an indicator of relative productivity. However, in some aspects of the invention, the expression level may be used to predict whether the milk-producing animal has high productivity or low productivity. High and low productivity are defined above.

[0130] High productivity may, for example, be predicted from a quantified expression that is less than a reference value. That is, high productivity may be predicted when quantified expression of one or more haemoglobin genes is less than a reference value for each of the one or more haemoglobin genes. The reference value may, for example, be the average quantified expression of the relevant haemoglobin gene for a population of milk-producing animals of known low productivity. The population is preferably a population of animals belonging to the same species as the milk-producing animal or parent whose productivity is being predicted. The average quantified expression may, for example, refer to the mean, median or mode quantified expression. Preferably, the average quantified expression refers to the mean quantified expression.

[0131] A decrease in quantified expression of one or more haemoglobin genes, relative to the reference value, may indicate that the milk-producing animal has high productivity. For example, a negative fold change in quantified expression of one or more haemoglobin genes, relative to the reference value, may indicate that the milk-producing animal has high productivity. No change or a increase in quantified expression of one or more haemoglobin genes, relative to the reference value, may indicate that the milk-producing animal has low productivity. For example, an insignificant fold change or a positive fold change in quantified expression of one or more haemoglobin genes, relative to the reference value, may indicate that the milk-producing animal has low productivity.

[0132] High productivity may, for example, be predicted from a quantified expression of HBA, HBA1, HBB, HBM and / or HBQ1 that represents a negative log2 fold change compared to the reference value. For instance, high productivity may be predicted from:

[0133] (1) a quantified expression of HBA (or an equivalent gene in a non-cattie bovid) that represents a log2 fold change of at least -0.8 compared to the reference value; such as at least -0.9, at least -1.0, at least -1.1, at least -1.2, at least -1.3, at least -1.4, at least -1.5, at least -1.6, -0.8 to -2.6, -0.9 to -2.5, -1.0 to -2.4, -1.1 to -2.3, -1.2 to -2.2, -1.3 to -2.1, -1.4 to -2.0, -1.5 to -1.9, -1.6 to -1.8, -1.0 to -2.0, -1.3 to -1.9, -1.6 to -1.7, about -1.6, about -1.65 or about -1.7. In one aspect of the disclosure, the quantified expression of HBA represents a log2 fold change of -1.0 to -2.0 (such as -1.3 to -1.9, -1.6 to -1.7, about -1.6, about -1.65 or about -1.7) compared to the reference value. In a particular aspect of the disclosure, the quantified expression of HBA (or an equivalent gene in a non-cattie bovid) represents a log2 fold change of about -1.65. In this context, the term “about” may mean the specified value ± 0.5.

[0134] (2) a quantified expression of HBA1 (or an equivalent gene in a non-cattie bovid) that represents a log2 fold change of at least -0.8 compared to the reference value; such as at least -0.9, at least -1.0, at least -1.1, at least -1.2, at least -1.3, at least -1.4, at least -1.5, at least -1.6, -0.8 to -2.6, -0.9 to -2.5, -1.0 to -2.4, -1.1 to -2.3, -1.2 to -2.2, -1.3 to -2.1, -1.4 to -2.0, -1.5 to -1.9, -1.6 to -1.8, -1.0 to -2.0, -1.3 to -1.9, -1.6 to -1.7, about -1.6, about -1.65 or about -1.7. In one aspect of the disclosure, the quantified expression of HBA1 (or an equivalent gene in a non-cattie bovid) represents a log2 fold change of -1.0 to -2.0 (such as -1.3 to -1.9, -1.6 to -1.7, about -1.6, about -1.65 or about -1.7) compared to the reference value. In a particular aspect of the disclosure, the quantified expression of HBA1 represents a log2 fold change of about -1.65. In this context, the term “about” may mean the specified value ± 0.5.

[0135] (3) a quantified expression of HBB (or an equivalent gene in a non-cattie bovid) that represents a log2 fold change of at least -0.7 compared to the reference value; such as at least -0.75 at least -0.8, at least -0.9, at least -1.0, at least -1.1, at least -1.2, at least -1.3, at least -1.4, at least -1.5, at least -1.6, -0.7 to -2.5, -0.8 to -2.4, -0.9 to -2.3, -1.0 to -2.2, -1.1 to -2.1, -1.2 to -2.0, -1.3 to -1.9, -1.4 to -1.8, -1.5 to -1.7, -1.5 to -1.6, or about -1.5. In one aspect of the disclosure, the quantified expression of HBB (or an equivalent gene in a noncattle bovid) represents a log2 fold change of -1.0 to -2.0 (such as -1.3 to -1.7, -1.4 to -1.6, or about -1.5) compared to the reference value. In a particular aspect of the disclosure, the quantified expression of HBB (or an equivalent gene in a non-cattie bovid) represents a log2 fold change of about -1.5. In this context, the term “about” may mean the specified value ± 0.5. In this context, the term “about” may mean the specified value ± 0.5.

[0136] (4) a quantified expression of HBM (or an equivalent gene in a non-cattie bovid) that represents a log2 fold change of at least -0.8 compared to the reference value; such as at least -0.9, at least -1.0, at least -1.1, at least -1.2, at least -1.3, at least -1.4, at least -1.5, at least -1.6, -0.8 to -2.6, -0.9 to -2.5, -1.0 to -2.4, -1.1 to -2.3, -1.2 to -2.2, -1.3 to -2.1, -1.4 to -2.0, -1.5 to -1.9, -1.6 to -1.8, -1.0 to -2.0, -1.3 to -1.9, -1.6 to -1.7, about -1.6, about -1.65 or about -1.7. In one aspect of the disclosure, the quantified expression of HBM (or an equivalent gene in a non-cattie bovid) represents a log2 fold change of -1.0 to -2.0 (such as -1.3 to -1.9, -1.6 to -1.7, about -1.6, about -1.65 or about -1.7) compared to the reference value. In a particular aspect of the disclosure, the quantified expression of HBM (or an equivalent gene in a non-cattie bovid) represents a log2 fold change of about -1.65. In this context, the term “about” may mean the specified value ± 0.5. In this context, the term “about” may mean the specified value ± 0.5.

[0137] (5) a quantified expression of HBQ1 (or an equivalent gene in a non-cattie bovid) that represents a log2 fold change of at least -0.10 compared to the reference value; such at least -0.11, at least -0.12, at least -0.13, at least -0.14, at least -0.15, at least -0.16, at least - 0.17, at least -0.18, at least -0.19, at least -0.20, at least -0.21, at least -0.22, at least -0.23, at least -0.24, at least -0.25, at least -0.26, at least -0.27, at least -0.28, -0.23 to -0.33, -0.24 to -0.32, -0.25 to -0.31, -0.26 to -0.30, -0.27 to -0.29, or about -0.28. In one aspect of the disclosure, the quantified expression of HBQ1 (or an equivalent gene in a non-cattie bovid) represents a log2 fold change of -0.23 to -0.33 (such as -0.25 to -0.31, -0.27 to -0.29, or about -0.28) compared to the reference value. In a particular aspect of the disclosure, the quantified expression of HBQ1 (or an equivalent gene in a non-cattie bovid) represents a log2 fold change of about -0.28. In this context, the term “about” may mean the specified value ± 0.05.

[0138] In each case, the log2 fold change may be a statistically significant log2 fold change. In each case, the animal may be any bovid, but is preferably a cow or heifer or a parent of a cow or heifer.

[0139] High productivity may, for example, be predicted from: (1); (2); (3); (4); (5); (1) and (2); (1) and (3); (1) and (4); (1) and (5); (2) and (3); (2) and (4); (2) and (5); (3) and(4); (3) and (5); (4) and (5); (1), (2) and (3); (1), (2) and (4); (1), (2) and (5); (1), (3) and (4);

[0140] (1), (3) and (5); (1), (4) and (5); (2), (3) and (4); (2), (3) and (5); (2), (4) and (5); (3), (4) and (5); (1), (2), (3) and (4); (1), (2), (3) and (5); (1), (2), (4) and (5); (1), (3), (4) and (5);

[0141] (2), (3), (4) and (5); (1), (2), (3), (4) and (5) . In one aspect of the disclosure, high productivity is predicted from (1), (2), (3), (4) and (5).

[0142] High productivity may, for example, be predicted from a quantified expression of HBA, HBA1, HBB, HBM and / or HBQ1 (or an equivalent gene in a non-cattie bovid) that represents a percentage decrease compared to the reference value. For instance, high productivity may be predicted from:

[0143] (A) a quantified expression of HBA (or an equivalent gene in a non-cattie bovid) that represents a decrease of at least 10% compared to the reference value; such as at least 20%, at least 25%, at least 30%, at least 50%, at least 60%, at least 70%, at least 75%, at least 80%, at least 90%, or at least 95%.

[0144] (B) a quantified expression of HBA1 (or an equivalent gene in a non-cattie bovid) that represents a decrease of at least 10% compared to the reference value; such as at least 20%, at least 25%, at least 30%, at least 50%, at least 60%, at least 70%, at least 75%, at least 80%, at least 90%, or at least 95%.

[0145] (C) a quantified expression of HBB (or an equivalent gene in a non-cattie bovid) that represents a decrease of at least 10% compared to the reference value; such as at least 20%, at least 25%, at least 30%, at least 50%, at least 60%, at least 70%, at least 75%, at least 80%, at least 90%, or at least 95%.

[0146] (D) a quantified expression of HBM (or an equivalent gene in a non-cattie bovid) that represents a decrease of at least 10% compared to the reference value; such as at least 20%, at least 25%, at least 30%, at least 50%, at least 60%, at least 70%, at least 75%, at least 80%, at least 90%, or at least 95%.

[0147] (E) a quantified expression of HBQ1 (or an equivalent gene in a non-cattie bovid) that represents a decrease of at least 10% compared to the reference value; such as at least 20%, at least 25%, at least 30%, at least 50%, at least 60%, at least 70%, at least 75%, at least 80%, at least 90%, or at least 95%.

[0148] High productivity may, for example, be predicted from: (A); (B); (C); (D); (E); (A) and (B); (A) and (C); (A) and (D); (A) and (E); (B) and (C); (B) and (D); (B) and (E); (C) and(D); (C) and (E); (D) and (E); (A), (B) and (C); (A), (B) and (D); (A), (B) and (E); (A), (C) and (D); (A), (C) and (E); (A), (D) and (E); (B), (C) and (D); (B), (C) and (E); (B), (D) and (E); (C), (D) and (E); (A), (B), (C) and (D); (A), (B), (C) and (E); (A), (B), (D) and (E); (A), (C), (D) and (E); (B), (C), (D) and (E); (A), (B), (C), (D) and (E) . In one aspect of the disclosure, high productivity is predicted from (A), (B), (C), (D) and (E).

[0149] Low productivity may, for example, be predicted from a quantified expression of HBA, HBA1, HBB, HBM and / or HBQ1 (or an equivalent gene in a non-cattie bovid) that represents a positive log2 fold change compared to the reference value. Low productivity may, for example, be predicted from a quantified expression of HBA, HBA1, HBB, HBM and / or HBQ1 (or an equivalent gene in a non-cattie bovid) that does not significantly differ from the reference value. For instance, low productivity may be predicted from:

[0150] (i) a quantified expression of HBA (or an equivalent gene in a non-cattie bovid) that represents a positive log2 fold change compared to the reference value, or which does not significantly differ from the reference value; (ii) a quantified expression of HBA1 (or an equivalent gene in a non-cattie bovid) that represents a positive log2 fold change compared to the reference value, or which does not significantly differ from the reference value;

[0151] (iii) a quantified expression of HBB (or an equivalent gene in a non-cattie bovid) that represents a positive log2 fold change compared to the reference value, or which does not significantly differ from the reference value;

[0152] (iv) a quantified expression of HBM (or an equivalent gene in a non-cattie bovid) that represents a positive log2 fold change compared to the reference value, or which does not significantly differ from the reference value; and / or

[0153] (v) a quantified expression of HBQ1 (or an equivalent gene in a non-cattie bovid) that represents a positive log2 fold change compared to the reference value, or which does not significantly differ from the reference value.

[0154] Low productivity may, for example, be predicted from: (i); (ii); (iii); (iv); (v); (i) and (ii); (i) and (iii); (i) and (iv); (i) and (v); (ii) and (iii); (ii) and (iv); (ii) and (v); (iii) and(iv); (iii) and (v); (iv) and (v); (i), (ii) and (iii); (i), (ii) and (iv); (i), (ii) and (v); (i), (iii) and (iv); (i), (iii) and (v); (i), (iv) and (v); (ii), (iii) and (iv); (ii), (iii) and (v); (ii), (iv) and (v); (iii), (iv) and (v); (i), (ii), (iii) and (iv); (i), (ii), (iii) and (v); (i), (ii), (iv) and (v); (i), (iii), (iv) and (v); (ii), (iii), (iv) and (v); (i), (ii), (iii), (iv) and (v) . In one aspect of the disclosure, low productivity is predicted from (i), (ii), (iii), (iv) and (v).

[0155] Low productivity may, for example, be predicted from a quantified expression of HBA, HBA1, HBB, HBM and / or HBQ1 (or an equivalent gene in a non-cattie bovid) that represents no change compared to the reference value, or a percentage increase compared to the reference value. For instance, low productivity may be predicted from:

[0156] (I) a quantified expression of HBA (or an equivalent gene in a non-cattie bovid) that represents no change compared to the reference value, or an increase of at least 10% compared to the reference value; such as at least 20%, at least 25%, at least 30%, at least 50%, at least 60%, at least 70%, at least 75%, at least 80%, at least 90%, or at least 95%.

[0157] (II) a quantified expression of HBA1 (or an equivalent gene in a non-cattie bovid) that represents no change compared to the reference value, or an increase of at least 10% compared to the reference value; such as at least 20%, at least 25%, at least 30%, at least 50%, at least 60%, at least 70%, at least 75%, at least 80%, at least 90%, or at least 95%. (III) a quantified expression of HBB (or an equivalent gene in a non-cattie bovid) that represents no change compared to the reference value, or an increase of at least 10% compared to the reference value; such as at least 20%, at least 25%, at least 30%, at least 50%, at least 60%, at least 70%, at least 75%, at least 80%, at least 90%, or at least 95%.

[0158] (IV) a quantified expression of HBM (or an equivalent gene in a non-cattie bovid) that represents no change compared to the reference value, or an increase of at least 10% compared to the reference value; such as at least 20%, at least 25%, at least 30%, at least 50%, at least 60%, at least 70%, at least 75%, at least 80%, at least 90%, or at least 95%.

[0159] (V) a quantified expression of HBQ1 (or an equivalent gene in a non-cattie bovid) that represents no change compared to the reference value, or an increase of at least 10% compared to the reference value; such as at least 20%, at least 25%, at least 30%, at least 50%, at least 60%, at least 70%, at least 75%, at least 80%, at least 90%, or at least 95%.

[0160] Low productivity may, for example, be predicted from: (I); (II); (III); (IV); (V); (I) and (II); (I) and (III); (I) and (IV); (I) and (V); (II) and (III); (II) and (IV); (II) and (V); (III) and(IV); (III) and (V); (IV) and (V); (I), (II) and (III); (I), (II) and (IV); (I), (II) and (V);

[0161] (I), (III) and (IV); (I), (III) and (V); (I), (IV) and (V); (II), (III) and (IV); (II), (III) and (V);

[0162] (II), (IV) and (V); (III), (IV) and (V); (I), (II), (III) and (IV); (I), (II), (III) and (V); (I), (II), (IV) and (V); (I), (III), (IV) and (V); (II), (III), (IV) and (V); (I), (II), (III), (IV) and (V) . In one aspect of the disclosure, low productivity is predicted from (I), (II), (III), (IV) and (V).

[0163] Method of using a trained machine learning algorithm model

[0164] The disclosure provides a method of using a trained machine learning algorithm model to characterise a milk-producing animal as being of high productivity or low productivity, comprising: (i) receiving input data that comprises data derived from a transcriptomic analysis performed on a sample obtained from the milk-producing animal, or from a parent of the milk-producing animal; and (ii) processing the input data using a trained machine learning algorithm model to generate an output representing information which characterises the milk-producing animal as being of high productivity or low productivity.

[0165] Categorisation of a milk-producing animal as high productivity or low productivity can be used to determine how the milk-producing animal, or a parent of the milk- producing animal, is managed. For instance, a high productivity milk-producing animal may be retained in the herd, or a low productivity milk-producing animal may be removed from the herd, with a view of improving productivity at the herd level. Productivity of the herd may also be improved by selecting high productivity milk-producing animals, or a parent giving rise to high productivity offspring, for use in breeding new herd members. Improvements in productivity have important implications for sustainable farming and the economic viability of the herd.

[0166] Productivity

[0167] Productivity is described above in connection with the method of predicting productivity. Any of the features described above for productivity may also apply to the method of using a trained machine learning algorithm model.

[0168] Milk-producing animal

[0169] Milk-producing animals are described above in connection with the method of predicting productivity. Any of the features described above for milk-producing animals may also apply to the method of using a trained machine learning algorithm model.

[0170] Sample

[0171] Samples are described above in connection with the method of predicting productivity. Any of the features described above for samples may also apply to the method of using a trained machine learning algorithm model.

[0172] Input data

[0173] The input data comprises data derived from a transcriptomic analysis performed on a sample obtained from the milk-producing animal, or from a parent of the milk-producing animal. A transcriptomic analysis is a study of an organism’s transcriptome, which is the sum of all its RNA transcripts. In the context of the method of using a trained machine learning algorithm model, the organism is the milk-producing animal, or the parent of the milk-producing animal. Numerous techniques for transcriptomic analysis are well-known in the art. These include RNA-Seq, microarrays, next generation sequencing and digital PCR, for example. Therefore, the transcriptomic analysis may be performed on the sample by or using RNA- Seq. The transcriptomic analysis may be performed on the sample by or using a microarray. The transcriptomic analysis may be performed on the sample by or using next generation sequencing. The transcriptomic analysis may be performed on the sample by or using digital PCR. Any other suitable transcriptomic technique may though be used.

[0174] The data derived from the transcriptomic analysis (and thus the input data) may, for example, comprise expression data for one or more genes comprised in a cell comprised in the sample. The data derived from the transcriptomic analysis may, for example, comprise expression data for multiple genes comprised in a cell comprised in the sample.

[0175] Preferably, the data derived from the transcriptomic analysis comprise expression data for at least 30%, at least 35%, at least 40%, at least 45%, at least 50%, at least 55%, at least 60%, 65%, at least 70%, at least 75%, at least 80%, at least 85%, at least 90%, at least 95%, at least 97%, at least 98% or at least 99% of the genes comprised in a cell comprised in the sample. The data derived from the transcriptomic analysis may, for example, comprise expression data for all of the genes comprised in a cell comprised in the sample.

[0176] Accordingly, the data derived from the transcriptomic analysis (and thus the input data) may, for example, comprise expression data for one or more genes comprised in the milk-producing animal or parent. The data derived from the transcriptomic analysis may, for example, comprise expression data for multiple genes comprised in the milk-producing animal or parent. Preferably, the data derived from the transcriptomic analysis comprise expression data for at least 30%, at least 35%, at least 40%, at least 45%, at least 50%, at least 55%, at least 60%, 65%, at least 70%, at least 75%, at least 80%, at least 85%, at least 90%, at least 95%, at least 97%, at least 98% or at least 99% of the genes comprised in the milk-producing animal or parent. The data derived from the transcriptomic analysis may, for example, comprise expression data for all of the genes comprised in the milk-producing animal or parent.

[0177] Expression data for a particular gene may indicate the amount of expression (at the mRNA level) of that gene in the cell and, therefore, the milk-producing animal or parent. The input data may comprise one or more individual data units. Each individual data unit may relate to the same type of data. Each individual data unit may, for example, be derived from or consist of expression data, such as mRNA expression data. In a first aspect, the input data may comprise plural individual data units, each individual data unit being derived from or consisting of expression data for a different gene. Essentially, the input data may comprise multiple units of expression data, each relating to a different gene in a cell comprised in the sample. In a second aspect, the input data may comprise data derived from or consisting of expression data representing a composite expression analysis containing expression data for multiple different genes. Essentially, the input data may comprise a single unit of expression data that relates to multiple different genes in a cell comprised in the sample. The first and second aspects are not mutually exclusive. That is, the input data may comprise (a) plural individual data units, each individual data unit being derived from or consisting of expression data for a different gene; and (b) data derived from or consisting of expression data representing a composite expression analysis containing expression data for multiple different genes.

[0178] The input data may comprise two or more individual data units each relating to a different type of data. For example, one of the two or more individual data units may be derived from or consist of expression data, and one of the two or more individual data units may not be derived from or consist of expression data. In other words, in addition to the data derived from or consisting of expression data, the input data may comprise data that is neither derived from nor consists of expression data. The data that are neither derived from nor consists of expression data may, for example, comprise any other kind of data relating to the milk-producing animal or parent, such as genetic data, clinical data or epidemiological data.

[0179] Trained machine learning algorithm model

[0180] The input data are processed using a trained machine learning algorithm model in order to generate an output representing information which characterises the milkproducing animal as being of high productivity or low productivity. The machine learning algorithm model may comprise or consist of a Random Forest or XGBoost algorithm, for example. Preferably, the machine learning algorithm model comprises or consists of a XGBoost algorithm.

[0181] The machine learning algorithm model may perform a single processing step on the input data in order to produce the output. Alternatively, the machine learning algorithm model may perform a plurality of processing steps on the input data in order to produce the output. One or more of the plurality of processing steps may produce a sub-output that is used by the machine learning algorithm model in the next processing step. That is, two or more of the plurality of processing steps may be performed in series. Two or more of the plurality of processing steps may be performed in parallel.

[0182] Any of the processing steps may comprise a classification step. The classification step may be a binary classification step that produces an output associated with one of two categories. For instance, the classification step may be a binary classification step that produces an output associated with (1) high productivity or (2) low productivity. High and low productivity are described above.

[0183] Output

[0184] The output of the trained machine algorithm model represents information which characterises the milk-producing animal as being of high productivity or low productivity.

[0185] The output may be of a discrete nature. For example, the output may be a classification that either the input data is associated with high productivity or the input data is associated with low productivity. High and low productivity are described above.

[0186] The output may be of a continuous nature. For example, the output may represent or consist of the probability that the input data is associated with high productivity. The output value may represent or consist of the probability that the input data is associated with the low productivity.

[0187] A continuous output may, for example, be numerical. A numerical cut-off may be applied in order to convert the continuous output to a discrete output. For example, an output value that (i) represents or consists of the probability that the input data is associated with high productivity and (ii) is greater than a particular numerical cut-off may be converted to a classification that the input data is associated with high productivity. An output value that (i) represents or consists of the probability that the input data is associated with high productivity and (ii) is less than a particular numerical cut-off may be converted to a classification that the input data is associated with low productivity. An output value that (i) represents or consists of the probability that the input data is associated with low productivity and (ii) is greater than a particular numerical cut-off may be converted to a classification that the input data is associated with low productivity. An output value that (i) represents or consists of the probability that the input data is associated with low productivity and (ii) is less than a particular numerical cut-off may be converted to a classification that the input data is associated with high productivity.

[0188] Method for training a machine learning algorithm model

[0189] The disclosure also provides a method of training a machine learning algorithm model to characterise a milk-producing animal as being of high productivity or low productivity, the method comprising training the machine learning algorithm model using a plurality of training data units, each training data unit comprising (i) input data that comprises data derived from a transcriptomic analysis performed on a sample obtained from a respective training milk-producing animal or from a parent of the respective training milk-producing animal, and (ii) data representing information about the productivity of the respective training milk-producing animal.

[0190] The machine learning algorithm model may therefore be provided with training data (plurality of training data units) from many milk-producing animals and / or parents for whom productivity is known. An aim of the training process is to enable the machine learning algorithm model to generate an output that accurately characterises milkproducing animal for whom productivity is not known as high productivity or low productivity. The training process may achieve this by iteratively adjusting parameters defining the machine learning algorithm model until the machine learning algorithm model is able to accurately characterise the training milk-producing animals as high productivity or low productivity. The training process may thus involve comparisons between outputs of the machine learning model for the training milk-producing animals and the corresponding known productivity of the training milk-producing animals.

[0191] The adjustment of parameters of the machine learning algorithm model may, for example, be performed by minimizing a loss function, for example using a gradient descent process. In some embodiments, a cross-validation process is performed. For instance, the plurality of training data units may be divided into k subsets. The machine learning algorithm model may be trained on all but one (i.e. k-T) of the subsets, and then evaluated on the subset that was not used for training. The process may be repeated k times, with a different subset reserved for evaluation (excluded from training) each time.

[0192] Training data units

[0193] A plurality of training data units is used to train the machine learning algorithm model. Each training data unit comprises (i) input data that comprises data derived from a transcriptomic analysis performed on a sample obtained from a respective training milkproducing animal or from a parent of the respective training milk-producing animal, and (ii) data representing information about the productivity of the respective training milkproducing animal.

[0194] Input data is described in detail above in connection with a method for using a trained machine learning algorithm model. Any of the aspects described above may also apply to the method for training a machine learning algorithm model.

[0195] Data representing information about the productivity of the respective training milk-producing animal may include data in relation to any characteristic indicative of high productivity. Such characteristics are described in detail above in connection with the method of predicting productivity of a milk-producing animal, and include milk yield, milk fat percentage, milk protein percentage, and milking speed for example. Any of the aspects described above may also apply to the method for training a machine learning algorithm model.

[0196] Machine learning algorithm model

[0197] Machine learning algorithm models are described in detail above. Any of the aspects described above may also apply to the method for training a machine learning algorithm model.

[0198] Outputs Outputs are described in detail above in connection with a method for using a machine learning algorithm model. Any of the aspects described above may also apply to the method for training a machine learning algorithm model.

[0199] Productivity

[0200] Productivity is described above in connection with the method of predicting productivity. Any of the features described above for productivity may also apply to the method of training machine learning algorithm model.

[0201] Milk-producing animal

[0202] Milk-producing animals are described above in connection with the method of predicting productivity. Any of the features described above for milk-producing animals may also apply to the method of training machine learning algorithm model.

[0203] Sample

[0204] Samples are described above in connection with the method of predicting productivity. Any of the features described above for samples may also apply to the method of training machine learning algorithm model.

[0205] Method of selecting a milk-producing animal for breeding

[0206] The disclosure provides a method of selecting a milk-producing animal for breeding. The method may comprise (i) performing the method of predicting the productivity of a milk-producing animal of the disclosure, and selecting the milkproducing animal or parent for breeding if the method predicts high productivity. The method may alternatively comprise (ii) performing the method of using a trained machine learning algorithm model to characterise a milk-producing animal as being of high productivity or low productivity of the disclosure, and selecting the milk-producing animal or parent for breeding if the method characterises the milk-producing animal as being of high productivity.

[0207] The purpose of the method is to identify individuals having high productivity, or giving rise to progeny having high productivity, such that those individuals can be included in a breeding pair. Inclusion of an individual having high productivity, or giving rise to progeny having high productivity, in a breeding pair may result in high productivity offspring from the breeding pair. Inclusion of an individual having high productivity, or giving rise to progeny having high productivity, in a breeding pair may increase the likelihood of high productivity offspring arising from the breeding pair.

[0208] The milk-producing animal is the milk-producing animal, or a parent of a milkproducing animal, from which sample is obtained in (i) or (ii). Thus, the milk-producing animal may be a bovid, such as a bovine or caprine. The bovine may be a Bos taurus. The caprine may be a Capra hircus or an Ovis aries. The milk-producing animal may be of any gender. For instance, the milk-producing animal may be female. For instance, the milk-producing animal may be a cow, heifer, doe, doe kid, ewe, gimmer or ewe lamb. Such milk-producing animals are described above. Any of the aspects of milk-producing animals described above may apply to the method of selecting a milk-producing animal for breeding. The milk-producing animal may be male. For instance, the milk-producing animal may be a bull, a buck or billy goat (i.e. a male goat), or a ram or tup (i.e. a male sheep).

[0209] Any of the aspects described above in connection with the method of predicting the productivity of a milk-producing animal or the method of using a trained machine learning algorithm model to characterise a milk-producing animal as being of high productivity or low productivity may also apply to the method of selecting a milk-producing animal for breeding.

[0210] Computer program

[0211] Computer programs may be provided that comprise instructions that, when executed by a computer system, instruct the computer system to perform the method of any of the aspects of the present disclosure.

[0212] The computer program may embody a method described herein as code (e.g. software code) and / or data. Such code and data can be stored on one or more computer- readable medium, which may include any device or medium that can store code and / or data for use by a computer system. When a computer system reads and executes the code and / or data stored on a computer-readable medium, the computer system performs the methods and processes embodied as data structures and code stored within the computer- readable storage medium. In certain embodiments, one or more of the steps of the methods and processes described herein can be performed by a processor (e.g., a processor of a computer system or data storage system). It should be appreciated by those skilled in the art that computer-readable media include removable and non-removable structures / devices that can be used for storage of information, such as computer-readable instructions, data structures, program modules, and other data used by a computing system / environment. A computer-readable medium includes, but is not limited to, volatile memory such as random access memories (RAM, DRAM, SRAM); and non-volatile memory such as flash memory, various read-only-memories (ROM, PROM, EPROM, EEPROM), magnetic and ferromagnetic / ferroelectric memories (MRAM, FeRAM), and magnetic and optical storage devices (hard drives, magnetic tape, CDs, DVDs); network devices; or other media now known or later developed that are capable of storing computer-readable information / data.

[0213] Apparatus and system

[0214] Data processing apparatus may be provided that comprises a processor configured to perform the method of any of the aspects of the present disclosure. To perform the method, the processor may execute the computer program mentioned above.

[0215] The data processing apparatus may further comprise an input device configured to receive the input data. The data processing apparatus may further comprise an output device configured to output the output generated by the machine learning algorithm model.

[0216] The data processing apparatus may, for example, comprise a milking machine.

[0217] The disclosure also provides a system for characterising a milk-producing animal as being of high productivity or low productivity, comprising: the data processing apparatus of an embodiment of the disclosure; and a transcriptomic analysis device configured to capture the input data. The transcriptomic analysis device may, for example, comprise a high-throughput sequencer capable of performing RNA-Seq. This would permit the system to both analyse the transcriptome of the milk-producing animal, and interpret the expression data obtained.

[0218] EXAMPLES The following Example is provided to illustrate the invention but are not intended to limit the invention.

[0219] Example 1

[0220] Design of the study is shown in Figure 1. In brief, phenotypic data were obtained in connection with 110 dairy cows from three farms, over a three-month period. The cows The cows were categorized into two groups based on their phenotype: good (n=66) and poor (n=44). Categorization was primarily based on milk yield, but also considered milk fat, milk protein, and milking speed.

[0221] Blood samples were collected from the cows, and high-quality RNA (RIN > 7, 280 / 260 ratio ~ 1.9-2.1) was extracted. This RNA was then used for RNA-seq analysis. In more detail, high-quality RNA libraries were prepared by Eurofins, Germany. The libraries were sequenced to a depth of approximately 30 million paired-ends reads per sample (Gold- standard). Prior to downstream analyses, meticulous pre-processing and quality control steps were applied to the raw RNA-seq data using FASTQC / MULTIQC. These steps included verifying MD5 checksums, trimming reads to remove low-quality bases or adapters, filtering out reads with low-quality scores, and aligning the reads accurately (with alignment rates of >85-100%) using STAR (as shown in Figure 2). For each sample, raw read counts, VST- normalized read counts, and log files were generated. The final output for each sample was a sorted and indexed bam file, approximately 5GB in size. Throughout the entire process, adherence to FAIR principles ensured that the data analysis was reproducible and transparent.

[0222] A comprehensive differential expression analysis was conducted to identify genes with significant differences in expression levels between good and poor cows. Potential confounding factors were identified and accounted for in the design model including reproduction status, age, batch and lactation days to ensure the reliability of the analysis. A stringent adjusted (FDR; false detection rate) p-value threshold of <0.05 was used to establish statistical significance.

[0223] Differentially expressed genes were subjected to pathway enrichment analysis using widely accepted bioinformatics tools. Established pathway databases, including Gene Ontology or Kyoto Encyclopaedia of Genes and Genomes (KEGG), and enrichR (which covers multiple databases) were utilized to identify biological pathways that were significantly enriched among the significantly differentially expressed genes. Pathway enrichment analysis aimed to provide functional insights into the underlying biological processes associated with the observed differences in productivity and, particularly, milk yield.

[0224] In addition, a training and test dataset derived from the RNA seq data was used to develop a predictive model that can effectively distinguish between animals with good and poor productivity. XGBoost and Random Forest models were trained. The XGBoost and Random Forest models are both popular machine learning algorithms that are known for their ability to handle complex datasets and make accurate predictions. These algorithms utilize an ensemble learning approach, which combines multiple decision trees to make collective predictions. XGBoost emphasizes boosting, an iterative process that assigns higher weights to misclassified samples, while Random Forest employs a combination of decision trees and feature randomness to improve prediction accuracy.

[0225] Results and discussion

[0226] Differential gene expression in good and poor cows

[0227] RNA-seq found that numerous genes (n=262) were significantly different (FDR <0.05) between good and poor cows. Pathway enrichment analysis identified multiple pathways significantly enriched (FDR <0.05) in these genes including those associated with nutrition-related processes directly linked to milk yield, as well as pathways associated with stress. Figure 3 highlights pathways that exhibited significant enrichment, providing potential targets for further investigation. Numerous genes identified as significantly different between cows with good and poor productivity have previously been linked to milk production (e.g. choline and alpha linolenic metabolism). However, no such link has been previously demonstrated for some of the other genes differentially expressed between cows with good and poor productivity, such as haemoglobin genes. Haemoglobin genes identified as significantly different between good and poor productivity cows included HBA, HBB, HBM and HBQ1. HBA and HBB were abundantly expressed across all samples with large fold-changes between good and poor cows. HBM and HBQ1 have low expression across all samples with large fold changes between good and poor cows. Raw read counts for haemoglobin genes are presented in Table 1. Violin plots and box plots showing expression in good and poor cows are provided in Figures 4 and 5 respectively. Fold changes are shown in Table 5.

[0228] Table 1: Raw read counts of HBA / 1, HBB, HBM and HBQ1 shown for 1 typical sample (in duplicate).

[0229] Table 2: Fold changes for HBA / 1, HBB, HBM and HBQ1 shown across all samples

[0230] Model predictive of productivity

[0231] Mixed results were seen for development of machine learning models had mixed results. In particular, the Random Forest model performed poorly, with only -62% accuracy when run on the test dataset. However, the XGBoost model achieved an accuracy of 78% in predicting whether a cow was a good or poor milk yi elder based on specific genes which each contributed different weights to the model, as shown in the confusion matrix and associated statistics below. Confusion Matrix and Statistics

[0232] Reference

[0233] Prediction no yes no 19 7 yes 7 32

[0234] Accuracy : 0,7846

[0235] 9® CI : C0.6651, 0.8769)

[0236] No Inf emotion Rate : 0,6

[0237] P- Value [Acc > NIR] : 0,801288

[0238] Kappa : 0,5513

[0239] Mcnemar’s Test P-Value : 1.000000

[0240] Sensitivity : @.73®8

[0241] Specificity : MM5

[0242] Pos Fred Value : 8,73®8

[0243] Keg Pred Value : 0.8205

[0244] Prevalence : 0.4000

[0245] Detection Rate : O.2923

[0246] Detection Prevalence : 0,4800

[0247] Balanced Accuracy : 0.7756

[0248] 'Positive' Class : no

[0249] This result suggests that the model could be used to effectively predict the productivity of dairy cattle based on RNA-seq data. This can be used to improve milk production within a herd, by adopting suitable management and breeding practices based on the categorisation of a milk-producing animal as good or poor.

[0250] SEQUENCE LISTING

[0251] SEQ ID NO: 1

[0252] CAGCCCCTGGGCCATAGCGCCCCATCTTATACACAGACAGACAGACACACACA

[0253] CTGTCAATCTCTCCTCGGCAGCACTGGGGCACGGATCTTAGCTATCAGTCCAGA

[0254] GCCTTTGCAAAGACAAGAGCCCAACCCTCCAAGCTCTCCAAACTCTGAATGAG

[0255] GATGGTTCCCTGGATGACAGTCTCAGGTATGTTCAGGGCTGGACACTCAGTGA

[0256] CCAAGTTTGAAGGGGGATGCTGGACACTAACCCCTCACCCAGACTGGGCAGAA

[0257] CCCCCCTCGGTCTCTGCCCTTTAGTGCGAAAGCAGCTTGCTTCACTTCGGAAAA

[0258] GATTGTGTTCCTCTCCCAGACTGGCAAACGGAAACCTCCATGCTACCCCAGCG

[0259] TGGGTACAAACTCGGTTCCAGGGGGTTTGGGTGGCTCAGAAAAGTGCGTGAGC

[0260] TGTTCAGCCCACAGAACCAGGGTAGAAGTTACCATAAATAGCTGCATTGCAGC

[0261] CAAAGCGGGACAGGGCAGGGGCCATCTTGGGACCCCGACACCCCACACACTC

[0262] TCCAGACCCATCTTTCCAGGCCAGCACCTTCTGGCCTCGCGCCAGCCAATGAG

[0263] CGCAGCGCGGGCGGGCGTGCCCCTGGCGCCGGGAGCATAAAAGCTCGCGCAC

[0264] TCGCAGCCCCGCACTCTTCTGGTCCAGACCCAGACTCTGAGAGAACCCACCAT

[0265] GGTGCTGTCTGCCGCCGACAAGGGCAATGTCAAGGCCGCCTGGGGCAAGGTTG

[0266] GCGGCCACGCTGCAGAGTATGGCGCAGAGGCCCTGGAGAGGTGAGCACCGCA

[0267] CTTGCCCCGAGGGGACCGGGCCGCACGCCGGGCGCGTCCTTGTCCCGGGCCGC

[0268] TCGGCCTAAGCCCGGCTTTCCCGCCTCTTCACCCAGGATGTTCCTGAGCTTCCC

[0269] CACCACCAAGACCTACTTCCCCCACTTCGACCTGAGCCACGGCTCCGCGCAGG

[0270] TCAAGGGCCACGGCGCGAAGGTGGCCGCCGCGCTGACCAAAGCGGTGGAACA

[0271] CCTGGACGACCTGCCCGGTGCCCTGTCTGAACTGAGTGACCTGCACGCTCACA

[0272] AGCTGCGTGTGGACCCGGTCAACTTCAAGGTGAGCTCGCGGGCAGGGCCGGGA

[0273] CAGATCTGGGCTAGCGGGGCAGAGAATGCGGCGGCGCCCCCACCCAGCCCCC

[0274] GCCCCCCTGACGTCCCCTCTCTCGGCAGCTTCTGAGCCACTCCCTGCTGGTGAC

[0275] CCTGGCCTCCCACCTCCCCAGTGATTTCACCCCCGCGGTCCACGCCTCCCTGGA

[0276] CAAGTTCTTGGCCAACGTGAGCACCGTGCTGACCTCCAAATACCGTTAAGCTG

[0277] GAGCCTCGGCGACCCCTACCCTGGCCTGGAGCCCCCTTGCGCTCTGCGCACTCT

[0278] CACCTCCTGATCTTTGAATAAAGTCTGAGTGGGCTGCAGTGTCTGTAGCCTCGG

[0279] GTCTCTGTGTCCGCGAACCGGCCGGGGTGGGCGTGGTTCTCAGTCCCTAAGAG TGAGAGGGTGGAGGAGGGCGGGGAGCTAAGGCTGAGGGTCCCAGAATCTGCT

[0280] GAACCAAGTTCCCCTCCTGGGAAGACTTTCAAGGGTTCTCTCTGAGGTGGGGA

[0281] GTGCTGAAATGGCCACCCTGTTTTGAAATTTTTCTGATCCCCCTTAAACATAGT

[0282] TAGATGAAAACAAAAGCAAACCACACAAATAGGCTCTTCCTTGTGCAAAAGTA

[0283] GGGGCAGGACAACCTCCAGGAGATTCCACTTCCTCCCTATCAGAGACAAGGGG

[0284] GATCGGAACAGGGATCGCGGCCCAGGCACCCGTCCTGGGTCACAGGAACTCTC

[0285] TGCGGCGTTGCGGGGGAGGGGCAGTCAGTGAGAAGAGGGGCCTCTAGCTCGC

[0286] TGCCGTCTTCAGCAGGCTCCTCACTCCCTTGTCCCTGGAGTCTGCGAAGCTCAC

[0287] TGTGCCCCTAGCCCTGCGTTCCCTGGCCTCTAAGCCCCCTCCCCACATCCATCC

[0288] GGGCTTCTGGGGTCTCCTGGCAGTTCTGCCTTGGTCCTGTGAGCAACCTTCAG

[0289] SEQ ID NO: 2

[0290] MVLSAADKGNVKAAWGKVGGHAAEYGAEALERMFLSFPTTKTYFPHFDLSHGS

[0291] AQVKGHGAKVAAALTKAVEHLDDLPGALSELSDLHAHKLRVDPVNFKLLSHSLL

[0292] VTLASHLPSDFTPAVHASLDKFLANVSTVLTSKYR

[0293] SEQ ID NO: 3

[0294] TATCTGTTCGTTCCAAGACTGCCACTTAGTTGCACTCAAAAACTCTCTGCGGAA

[0295] AGAAAGGAATCTGGCGCCAAGGCTGAAGAGGAACAGGAGGCTTGGACTGGGG

[0296] TGGGGGTGGGGTGGGGAGGAATTCGCTGATTACATGTGAACGGTGAGCCAAGT

[0297] GTGTTACTGGCTGCCTCTGGCATGGACTAGGCGCACTCAATCGCCCGTTCATTC

[0298] ACTGATACTGCCCAAGTTTAAAATGCCCAGAGTGTGCCAAGCCTAGGTCCGGG

[0299] GTGGATAGCACGGGCTGACTTACTCCCTTCCGTTCTCAAGACAGCTGACGGAC

[0300] TCTTACAGGATGCAGGAGCGGGCATCCACCCAGCTCCACAATCCCGCCCCTGC

[0301] CACCTGGCGCGAGGCTACCACTTTCGGGGAAGGTGGACGCAGCGGGCGGGAA

[0302] GCAGACGGTGGAAGCAAGAACCCCGGGTCAGAGTCCAGGTCTCCGTGGGTGG

[0303] GTGAGGGAAGCACCCATCGCCCGGCCGGGCGCAGGTCGGACGCCGCGCGCCC

[0304] TCTGCGGTCCAGGTCCGGCCGCGCATGCCGCGTGCCAGCCAATGAGCGCAGCG

[0305] CGGGCGGGCGTGCCCCTGGCGCCGGGAGCATAAAAGCTCGCGCACTCGCAGC

[0306] CCCGCACTCTTCTGGTCCAGACCCAGACTCTGAGAGAACCCACCATGGTGCTG

[0307] TCTGCCGCCGACAAGGGCAATGTCAAGGCCGCCTGGGGCAAGGTTGGCGGCC ACGCTGCAGAGTATGGCGCAGAGGCCCTGGAGAGGTGAGCACCGCACTTGCC

[0308] CCGAGGGGACCGGGCCGCACGCCGGGCGTGTCCTTGTCCCGGGCCGCTCGGCC

[0309] TAAGCCCGGCTTTCCCGCCTCTTCACCCAGGATGTTCCTGAGCTTCCCCACCAC

[0310] CAAGACCTACTTCCCCCACTTCGACCTGAGCCACGGCTCCGCGCAGGTCAAGG

[0311] GCCACGGCGCGAAGGTGGCCGCCGCGCTGACCAAAGCGGTGGAACACCTGGA

[0312] CGACCTGCCCGGTGCCCTGTCTGAACTGAGTGACCTGCACGCTCACAAGCTGC

[0313] GTGTGGACCCGGTCAACTTCAAGGTGAGCTCGCGGGCAGGGCCGGGACAGATC

[0314] TGGGCTAGCGGGGCAGAGAATGCGGCGGCGCCCCCACCCAGCCCCCGCCCCCC

[0315] TGACGTCCCCTCTCTCGGCAGCTTCTGAGCCACTCCCTGCTGGTGACCCTGGCC

[0316] TCCCACCTCCCCAGTGATTTCACCCCCGCGGTCCACGCCTCCCTGGACAAGTTC

[0317] TTGGCCAACGTGAGCACCGTGCTGACCTCCAAATACCGTTAAGCTGGAGCCTC

[0318] GGCGACCCCTACCCTGGCCTGGAGCCCCCTTGCGCTCTGCGCACTCTCACCTCC

[0319] TGATCTTTGAATAAAGTCTGAGTGGGCTGCAGTGTCTGTAGCCTCGGGTCTCTG

[0320] TGTCCGCGAACCGGCCGGGGTGGGCGTGGTTCTCATTCATTGCCTCGGACTAA

[0321] GGAGCTCTCAGGCAGAGGGAGAGAAGGGGAAAATGGGACGGAGGGGTGGGG

[0322] GTGGAGCCTGTCCCACGGCCACTGCCTGGGATTCTCTGGGCAGCCCTCACCCTC

[0323] AGCCTGGAGTGATTTCTGAGTACATTGGCCCTTCCCTGAATTCATCCCAGAGCC

[0324] TTGCATGTTCATCACGGAGATTTTTGCGCAAATGATCAATTTGTTGGCGGACCT

[0325] CCAGCCAGGACTCGGAGCTCTCCCGGGGGCTCCCGGAGCTGGGTGGTGGGGA

[0326] GAACACAGAGGCCCTCTGGGTGCCTCCCTCCCCACCAAGTGCTTAGAGGAAGG

[0327] TAGGGGTGCGTAGTGCCCTTGTAGAGCCAGGCCAGGTCCTGGAGGAACCTCTG

[0328] TCACCTGGTACACCCACCCCTCCAGGACACCCACCGGGCGATTGAGGGTCTGA

[0329] AAGCCTGATTTTGGGCACCCAGCTTAGGATCCACTGCACTGCCCTGAGGCCCA

[0330] CTGACTCCCCCTGGCGCTGCTGTGGAGAGGAGAGCCAAGGCCAGGGT

[0331] SEQ ID NO: 4

[0332] MVLSAADKGNVKAAWGKVGGHAAEYGAEALERMFLSFPTTKTYFPHFDLSHGS

[0333] AQVKGHGAKVAAALTKAVEHLDDLPGALSELSDLHAHKLRVDPVNFKVSSRAGP

[0334] GQIWASGAENAAAPPPSPRPPDVPSLGSF

[0335] SEQ ID NO: 5 TGGGGTTATGATTGATTTTTCATTTATTTGTATTTCTTACTACATGGAGAGGTCT

[0336] CTATGTTTTCACATATGGATTAATGGAGTAAGTACTTCACAGGACCTCTCGATC

[0337] CATACTTATTCTATTTTTCTATCATTATTTATTTGGCCATAAATTAAATTAAGAA

[0338] AAAATTAAATGAGTAGATAAGCAAATGAATATTTGTTTTCATACCAGAACGAT

[0339] TTAATCCCAAACAAGCAAACAAAAGAGATGCGTATTTAGAACAAGGGCAGAG

[0340] GTTGTATCCAGGTTGTACTTGTAATTCTTTTGCATATTCTGAAGGCACAGGAGA

[0341] TGATCCATCCACATAGTCTTAAGTCGAATCATAGTGGACAAATCCTTTCCATTT

[0342] TCTGGAGCCCGGAGTCTTCATTTGTGTAATAAGAAAATTGAGGAGGTGGTTTC

[0343] CAAGAGGTTACCTTCGTTGTGATTCTAAAATCTCTACAAGCAAACTTCCTAAGG

[0344] AAGATGACTTTAATAGCAATTTGTATTGCTGGAATGACTGAGATCTAGAGATG

[0345] CCCAGAAAGAGGGCTGATGGTCTAAAGTCAGTGCCAGGAAGACCAAGGAGAG

[0346] CTATGACTATCATCGTTCAAGCCTCACCCTGTGGAACCACAACTTGGCATGAGC

[0347] CAATCTGCTCACAGAAGCAGGGAGGGCAGGAGGCAGGGCTGGGCATAAAAGG

[0348] AAGAGCTGGGCCAGCTGCTGCTTACACTTGCTTCTGACACAACCGTGTTCACTA

[0349] GCAACTACACAAACAGACACCATGCTGACTGCTGAGGAGAAGGCTGCCGTCA

[0350] CCGCCTTTTGGGGCAAGGTGAAAGTGGATGAAGTTGGTGGTGAGGCCCTGGGC

[0351] AGGTAGGTATCCCACTTACAAGGCAGGTTTAAGGAGAGTGAAATGCACCTGGG

[0352] CGTGTGAGGACAGAGCCGTCCCTGAGATTCAGAGAGCTGCTGGCTTCCTCTGA

[0353] CCTTGTGCTGTTTTCTCCCCCTAGGCTGCTGGTTGTCTACCCCTGGACTCAGAG

[0354] GTTCTTTGAGTCCTTTGGGGACTTGTCCACTGCTGATGCTGTTATGAACAACCC

[0355] TAAGGTGAAGGCCCATGGCAAGAAGGTGCTAGATTCCTTTAGTAATGGCATGA

[0356] AGCATCTCGATGACCTCAAGGGCACCTTTGCTGCGCTGAGTGAGCTGCACTGT

[0357] GATAAGCTGCATGTGGATCCTGAGAACTTCAAGGTGAGTTTGTGGAATCCTCA

[0358] GTGTTCTCCTTCTTCTTTTTATGGTCAAGCTCATGTCATGAGGAGAAAGCTGAA

[0359] TGGCAGGACACAGTTTAGAATGGAGAAGAGGTATTCTGGTTAGATTACTAAGG

[0360] ACTCCTCAGAACCGTTTAGACTCTTTTAACCTCTTTGTTCACAACCAGTATTTCC

[0361] TCTGATTCATTCTTGTTCTCTGTTGTCTGCAATGTCCTCTTTTTAATTATATTTTT

[0362] TATTTTGAGGGTTTAATTTGAAAAAAAAATTATATATCAACTTTAAAAATTGTA

[0363] TCTAATATTTCCCCCTTATCTGTTTCTTTCAAGGAATAAAATGTTCTATTGCTTT

[0364] TTGAAATGATTCAAAATAATAAAAATAATAACAAGTTCTGGATTAAGTTAGAA

[0365] AGAGAGAAACATTTCTAAATATATATTCGGGAAGATATAGGTAGATTCACATC AGTAGTAACAACTTCACTTCAGTCATCTTTGTGCTTATATCCTACGGTCACAGC

[0366] TTGGGATAAGACTGAAATACCCTGAATCTAACCTTGGATTTCCCTCATAGCTCA

[0367] GTTGGTTAAGCATCTGCCTGCAATGCAAGAGATCCCAGTTCGATTCCTGGGTCG

[0368] GGAAGGATGGCTGGAGAAGGGATAGGCACCCACTCCAGTATTCTTGGGTTTCC

[0369] CTTGTGGCTCAGCTGGTAAAGAATCTGCCTGCAATGTGGGAGACCCAGCTTCT

[0370] ATCCCTGAGTTTGGAAGATCCCCTGGAGAAGGGAAAGGCTACCCACTCCAGTA

[0371] TTCTGGCCTGGAGAAATCTATGGACTGTAGAGTCCATGGGGTTGCAAAGAATC

[0372] AGACACGATTGAGAGACTCTCACTTCACTCACCTGCACTAACCCTGCCCTTGCT

[0373] TAATGTCTTTTCCACACAGCTCCTGGGCAACGTGCTAGTGGTTGTGCTGGCTCG

[0374] CAATTTTGGCAAGGAATTCACCCCGGTGCTGCAGGCTGACTTTCAGAAGGTGG

[0375] TGGCTGGTGTGGCCAATGCCCTGGCCCACAGATATCATTAAGCTCCCTTTCCTG

[0376] CTTTCCAGGAAAGGTTTTTTCATCCTCAGAGCCCAAAGATTGAATATGGAAAA

[0377] ATTATGAAGTGTTTTGAGCATCTGGCCTCTGCCTAATAAAGACATTTATTTTCA

[0378] TTGCACTGGTGTATTTAAATTATTTCACTGTCTCTTACTCAGATGGGCACATGG

[0379] GAGGGCAAAACACTGAAGACATAAAGAAATGAAGGGCTAGTCGAGACCTTGA

[0380] GAAAATATATCAGTATCTTGGACCCCATGACAGCAGTGGTTGTAAATAGCTGA

[0381] TGTTATGGAAAACAGGCTTTGCTCCTTAGCCTTACTCTCCCTTAAAGAATTCAA

[0382] GTTGCAGCTTGATTTGGGAGTTAGATATTGCTAAATTTTATTTAAATAAATTAT

[0383] GTTATTTACCCTTTCTTGTAAATGTCTTCTTTCTCTTTAATTATCCAGAATATCA

[0384] CTTAGATTCATTAAGTTCTTCTGCCTAATTACACCACTGTTTTAAGATTTTCTTT

[0385] TTGCATTTTACTGTCCCCATTTCTCTTTCTCCCCTGTTCTCTTTTTATCCTAGTTT

[0386] TCTCTGCCATCTTAAAAGAACTACAAGAAAGACAGAACCTTCTGTGCTGGATT

[0387] CGGAAAATGACATATGAATTATGAGTAATCCTTGTTCCTTCTTGCATCCTAATT

[0388] CTGAATCTCAGTTCAGTTCACTTCAGTCTCTCAGTTATGTACGATTCTTTGCAAT CCCATGGACT

[0389] SEQ ID NO: 6

[0390] MLTAEEKAAVTAFWGKVKVDEVGGEALGRLLVVYPWTQRFFESFGDLSTADAV

[0391] MNNPKVKAHGKKVLDSFSNGMKHLDDLKGTFAALSELHCDKLHVDPENFKLLG

[0392] NVLVVVLARNFGKEFTPVLQADFQKVVAGVANALAHRYH SEQ ID NO: 7

[0393] CACCCAGGTCCCCACAATCACGTGGACGGAGGGAGGGAGGGAGGGAAGGAAG

[0394] GAATGACTTGGAGACGGGCGGTCAGGGTGAGGTGGCCGGGCAGTACTGAGCA

[0395] GCGGCTACCGCAAGAGTCACCCAGAGCGGTCACTAACTCGCCCGCCCGCCCGC

[0396] CGGCTCTTCTCCTGCTGCCCGCCGGCTCTTCTCCTGCTACCCGCAGGCCAAGAC

[0397] CTACTTCCCTCACTTCGACCTGCACACGGGCTCCGCGCAGCTGCGCGCGCACG

[0398] GCTCCAGGGTGGTGGCCGCCGTGGGCGACGCGGTCAAGAGCATTGACAACGTG

[0399] ACGAGCGCGTTGTCCAAGCTGAGCGAGCTGCACGCCTACGTGCTGCGCGTGGA

[0400] CCCGGTCAACTTCAAGGTGGGTGCGGGGCCCCCGGCGCGGGTTTTGGGCCGGG

[0401] GCCCCGGGGGCTGGAGGAGGTCTCGGTGGGCGGGGTTTCGGGGCGGGGCGAG

[0402] GCCTCGGGGGCGGGGCTGCTCCCTGGGCCGGGGTCCCGGAGGGCGGGGCGAG

[0403] GTCGCGGAGGCGGAGCTTCGTCCGGGCCGCCAGCGCTGACCGCTCTCCCTCCC

[0404] GCCCCAGTTCCTGTCCCATTGCCTGCTGGTCACGTTGGCCTCGCACTTCCCCGC

[0405] CGACTTCACGGCCGACGCGCACGCCGCCTGGGACAAGTTCCTGTCGATCGTGT

[0406] CCGGCGTCCTGACGGAGAAGTACCGCTGAGGCCGTCGCAGGACCCCTAGAAC

[0407] AGGCTTCGAGCCCTCTTCTTCCCCTGCACCCTATTAGTGTTTCATAGGGACCCC

[0408] CAATAAATGGATGAGGATGGAAGGAGCCTGGTCGTGCGTTCTCAGTATTGGGG

[0409] GAAAGAAGGGAGAGTGACGGGAGAGAATGGGGACTTCGCAGGAACGCAGCCC

[0410] AACACGGTGTCACCCCCCGGCCTGGAGTGACTCCAAGGGGCGCTTTGGAGCGG

[0411] TCCTTGGCCTGCAGAGCCCAGAGTCGCTACCGAAGGGAGCAGGGCAGCCCAG

[0412] GACCGTCCACGCCGCGTTCTCAGGTCAGACCTGCACTGGGACCTCTGTCCCAC

[0413] CCTATGTCACCCAGTCTGTGGCTTCACTCGGCGCGGTGGGTACGCGTCGGGACT

[0414] CAGAGTACTTCCCTCAGAGTTTGTCATCCCAGCTCACCGCTGCGAGGACCCTTG

[0415] AAACCTCCTGGCCCCTCGGATTAAGGGGGCTCCCTGCCCCATGCGCTTTCTCCC

[0416] TGTCCCTCCATTGGCTCTGTCTGGCGGAGCAGTAACTTTGCCTTCAGGGGAATA

[0417] ATTGCAGACCCATCTAAAGTCTGCAGCAAAGGCACCCACACACGGCGTGATTT

[0418] AGTTTCTCCGCGGTGAGGAGAACCTGGCTTGGCGACCTGGGTCGCCAGGTTCT

[0419] CGCCGCGCTCCTGAAGATAAGGGCGGGCGGGGCGCCACGAGGGGGCCGCTAT

[0420] AAGGAGGCCGGGGCGGCTGGCGCGGCCCCTAAAGCACGCCACGCCAGCGTCA

[0421] TGCTCAGCGCCCAAGAGCGCGCCCACATAACACAGGTCTGGGACCTGATAGCC

[0422] GGACACGAGGCGCCCTTCGGGGCGGAGCTGCTGCGCAGGTGGGTTGGGGCGG GGTCTCCAGAATCGCTGGGAGGAGGAGACCAGGGTACTGGGACTGAGTCCCG

[0423] GGTTGGTGGCGAGTCTGAGCGGGTTTTCCCGGGGTCTGCGCAGCGCAGACGCT

[0424] GGAGGGACCGTCCTCCCCACATCCCTGCTCCCGCAGGCTCTTCACTGTGTACCC

[0425] TAGCACCAAGGTCTACTTCAGACATCTGGGTGACCACCCAGACGAGGTGCAGC

[0426] TGCTGAGCCACGGTCAACGCATGCTCCAGGCGGTGGGCGTGGCCGTGCAGTAC

[0427] ATGGACAACCTGCGCGCAGTCCTGAGCCCGCTTGCCGACCTGCACGCACAAGT

[0428] GCTGCGCGTGGACCCCACCAACTTCCCAGTGAGCCACCCGGGGCGCAGCAGGG

[0429] CTCGCGGCGCAGCCTAGGCGGGAGGGGAGGCGTTCGGGGAGGCACCGGCGAG

[0430] GACAGGGCCCTTCTCACCGGCCTCTTCTCCGGCAGCTGGTGATCCAGTGTTTCC

[0431] AGGTGGTGCTGGCCTCCCACCTGCAGGGCGAGTTCACGGTGGAGATGCAGGCG

[0432] GCGTGGGATAAGTTCCTGACGGGCGTGGCCGTGGTGCTGACGGAGAAGTACCG

[0433] CTGAGTTCCGTGCCCCGCGGGTCTAGGTCTGCGAAATTCAATAAACAGGCTCA

[0434] GTAGGCTGGAACCTGCGCACGTGTGTTCTTTGGGGAGCAGACGGACTGGCGTC

[0435] GAAGTCGCCGGTCGGAGAAGCTCCGGGTCCTGAGGGGAGACTCCAGGAACCG

[0436] TAGGGTGCAACACCGAATCTTTGCTGACAGTGGAACCAGGGGCCTAGGAGCAG

[0437] GGCCTCTGGCACTCAACACCGCCTGAATGAGGACATTTCCAGGGCTGACAGTC

[0438] CTGGGTCTGGAACGCGCCTGGCACCCTGAGACAGTTGCCCTCCTGGGGCAGCG

[0439] TCATGTGGTGGGCCCATGGAGTACTCCCTTTCGGCATCGCGGCTGCCAGACAC

[0440] ATCCTCATTCACTCTCATCTCCATTGCTAGCTCCCTCAGGAACTGCAAGTGTCA

[0441] ATCTTCCCCTGTTTAACACACAGTAGCGGGCACCAGAGACCCTCAACCAAGTT

[0442] TCTGGCCTGAGGTCAAACAGACTTGGTGGTCCAGGAATCCCTGGACCTCACCT

[0443] CTGTTCCTCCTCCATTTCTTCCAACTCCAGTCTCAGTCTCAGTGCCTGGCCCTCG

[0444] TCCGAATTAGTCCCTTCCTCTGAATCTCCTGGGGCAGAGAGTGGGGTATGTCGA

[0445] CACATTCCTAGTCCACATAGACA

[0446] SEQ ID NO: 8

[0447] MLSAQERAHITQVWDLIAGHEAPFGAELLRRLFTVYPSTKVYFRHLGDHPDEVQL

[0448] LSHGQRMLQAVGVAVQYMDNLRAVLSPLADLHAQVLRVDPTNFPLVIQCFQVVL ASHLQGEFTVEMQAAWDKFLTGVAVVLTEKYR

[0449] SEQ ID NO: 9 GTCTCTTAAGTCTCCTGCATTGGCAGGTGGATTCTTTACCATTAGCGCCAACTA

[0450] GATTTCTATTTGGGATGATGAGAAAGTTCTGGAATTAGTGGTGATGGTAACAC

[0451] ATAGTGTGAATGTAGTTAAAGACACTGAATTGTATACCTCAAATGGCTAAGAC

[0452] AGTAAATTTTATGTGAATTTCAATGCATTTTTTAAAAAAGGATGTATAGTGATG

[0453] CCTGGCACAGAATGAGTGCTTTATAGGTATTGGCTATTATTTCTAGGATCTATC

[0454] CTAGCTACTCTGAATGCTCATTGTCCTGAAAAGGGCATCTGCTATGAACCAGG

[0455] AACAACCTTGGGAAGAAGAAACTGTTACCTGCCTTATATTCCATAAATGAGAA

[0456] AACAGACTGACGTTCAATGCCAAACTCCATGCTAAAATTCAGGGCCACCTAAT

[0457] ATTGACACATACTAAATATCAGTTTCCAAAAATAGGGATTTCAGGGTCCAGGG

[0458] CGATTGGGAAGGCCCCAGGTTAGAGCTGACGCTCCGTGGAGAGCTCCTAGACC

[0459] CCTCCTGGCCAGATTCTGCGACCAGAGGCCCGGCGCCCCCTGGAGGCCAAGGG

[0460] GCCAGCAGGCCGAGAGGCGCAGGCGCAGCCCAGGCTTCACCGAGACGCCGAG

[0461] CGCGGGGCAGGCCGGTCTCCACCATGGTGCTGGCGCCGGCGGACCGCGCTGCA

[0462] GTGCTCGCGTTGTGGCGGAAGCTCGGGACCAACGTCGGCATCTACACGACGGA

[0463] GGCCCTGGAGAGGTGCGCACAGGTCCTGGCGCCTCAGGGCTCACCTGTCTTCA

[0464] GCGCGCTCGGGGGATGCGTCCCCCGCCGTCCTATTCCGGAGCCCTTCCTCCCCA

[0465] GGCCCCGCCATCAGGACGCGCCTCACTCGCCGTTCCTCCCGCAGGACCTTCGT

[0466] GGCCTTCCCTTCCTCCAAGACCTATTTCCTCCACCTAAACTTGAGCCCCGGCTC

[0467] CGCCCAGGTCGCAGCCCACGGCCAGAAGGTGGCCGACGCGCTGAGCTTGGCC

[0468] GTGAACCACCTGGACGACCTGCCGGGCACCCTGTCGTACCTGCGCGAGCTGCA

[0469] CACGCACAAGCTGCGCGTGGACCCGGTCTTTTTCAAGGTGAGCTCGCGGGTGG

[0470] GGCCGACGCCCATTGCCGGGGTAGGGGGCTCTCACCAGGCAAGGGGTGGTGC

[0471] ACCGCGTGGGAACCATTCTGTCTGCAGCTGCTGTGCCACTGCCTGCTGGTGACC

[0472] CTCGCCCGGCACTACCCCGGAGACTTCAGCCCCAACATGCACGCCTCGCTGGT

[0473] CAAGTTTCTGAACCACGTGATCTCGGCGCTGGCCCCCAGCAGTGGCTAAATTG

[0474] GGAGGGTGGTAGCGGGGGTCTCCAGTGTTTGAGTAAAGTCCGCAGTCCCGGCT

[0475] GTGCCCGTGTTCTCTTCTAGGGGCCTGACGAGGGTGGGACGCCCTTCCCAAAG

[0476] AGAGGCGGGAAGTCTGGACAGGGGTTCTCCACTCCCACGTGGTAAGCCCCGAC

[0477] GCACCATGCCCTTGAGCGTGGCCACTTAGAGGTGTCGGCTGGGCCCCCGCCCG

[0478] GACCGTCTTTCCCGCTGCAGCCTTCGGACACCTCACATGCACCTAGTTAGTACC

[0479] GCCCGGCACAGCTTGCGAGGAAATTAGGGACGGGCGCCCCTGCCGGGGTGGA GCGGGACTGCACCTCCTGAACTGAACGCGCAGAGCACTTTCAGCAACCGGGAA

[0480] GCTTGGGGTCAACGGTGCCTTTGCCAGACTCCGGACCTGGTCGGAGGTGGCGG

[0481] CCCCTGCAGCGAACCCTGCCCCCATCCCACACACCAGACAGAACTCAGAGCGT

[0482] CTTCCTTTAAAAATTTTTTCTCATGGAATCAATCACGGTCATCGCAGAAACTTC AGGAAACGACTCCCAATTGCTGCACCTCAACTGCCCATGCCTTCCTGTGGGTTC

[0483] CGGACAT

[0484] SEQ ID NO: 10

[0485] MVLAPADRAAVLALWRKLGTNVGIYTTEALERTFVAFPSSKTYFLHLNLSPGSAQ VAAHGQKVADALSLAVNHLDDLPGTLSYLRELHTHKLRVDPVFFKLLCHCLLVTL

[0486] ARHYPGDFSPNMHASLVKFLNHVISALAPSSG

Claims

Claims1. A method of predicting the productivity of a milk-producing animal, comprising:(i) quantifying the expression of one or more haemoglobin gene in a sample obtained from the milk-producing animal, or from a parent of the milk producing animal; and(ii) predicting productivity based on an inverse relationship between quantified expression and productivity.

2. The method of claim 1, wherein productivity is a measure of one or more of milk yield, milk fat percentage, milk protein percentage and milking speed.

3. The method of claim 1 or 2, wherein productivity is a measure of milk yield.

4. The method of any one of the preceding claims, wherein the milk-producing animal is a bovid, optionally a bovine or a caprine.

5. The method of claim 4, wherein the bovine is a Bos Iannis. optionally a cow or heifer.

6. The method of claim 4, wherein the caprine is a Capra hircus. optionally a doe or a doe kid.

7. The method of claim 4, wherein the caprine is an Ovis aries. optionally a ewe, ewe lamb or gimmer.

8. The method of claims 4 to 7, wherein the bovid is a dairy breed.

9. The method of any one of the preceding claims, wherein expression is quantified by measuring expression of the mRNA or protein encoded by the haemoglobin gene.

10. The method of any one of the preceding claims, wherein expression is quantified by RNA-Seq, a microarray, qPCR, Northern blot, next generation sequencing (NGS), or digital PCR.

11. The method of any one of the preceding claims, wherein the one or more haemoglobin gene comprises HBA, HBA1, HBB, HBM and / or HBQ1.

12. The method of any one of the preceding claims, wherein the sample is a blood sample, a milk sample, a hair sample, a saliva sample, a tissue sample or a semen sample.

13. The method of any one of the preceding claims, wherein high productivity is predicted from a quantified expression that is less than a reference value.

14. The method of claim 13, wherein high productivity is predicted from:(a) a quantified expression of HBA that represents a log2 fold change of at least -0.8 compared to the reference value;(b) a quantified expression of HBA1 that represents a log2 fold change of at least -0.8 compared to the reference value;(c) a quantified expression of HBB that represents a log2 fold change of at least -0.7 compared to the reference value;(d) a quantified expression of HBM that represents a log2 fold change of at least -0.8 compared to the reference value;(e) a quantified expression of HBQ1 that represents a log2 fold change of at least -0.14 compared to the reference value.

15. The method of claim 13 or 14, wherein the reference value is the average quantified expression for a population of milk-producing animals of known low productivity.

16. A method of using a trained machine learning algorithm model to characterise a milk-producing animal as being of high productivity or low productivity, comprising:(i) receiving input data that comprises data derived from a transcriptomic analysis performed on a sample obtained from the milk-producing animal, or from a parent of the milk-producing animal; and(ii) processing the input data using a trained machine learning algorithm model to generate an output representing information which characterises the milk-producing animal as being of high productivity or low productivity, wherein the data derived from the transcriptomic analysis comprises expression data for one or more haemoglobin genes.

17. A method of training a machine learning algorithm model to characterise milkproducing animal as being of high productivity or low productivity, the method comprising training the machine learning algorithm model using a plurality of training data units, each training data unit comprising (i) input data that comprises data derived from a transcriptomic analysis performed on a sample obtained from a respective training milkproducing animal or from a parent of the respective training milk-producing animal, and (ii) data representing information about the productivity of the respective training milkproducing animal, wherein the data derived from the transcriptomic analysis comprises expression data for one or more haemoglobin genes.

18. The method of claim 16 or 17, wherein productivity is a measure of one or more of milk yield, milk fat percentage, milk protein percentage and milking speed.

19. The method of any one of claims 16 to 184, wherein the milk-producing animal is a bovid, optionally a bovine or a caprine.

20. The method of any one of claims 16 to 19, wherein the transcriptomic analysis comprises RNA-Seq, a microarray, next generation sequencing or digital PCR.

21. The method of any one of claims 16 to 20, wherein the sample is a blood sample, a milk sample, a hair sample, a saliva sample, a tissue sample or a semen sample.

22. The method of any one of the preceding claims, wherein:(a) the sample is obtained from the milk-producing animal, which is an existing milkproducing animal; or(b) the sample is obtained from a parent of the milk-producing animal, wherein (i) the milk-producing animal is the potential progeny of the parent and (ii) the parent is the prospective sire or dam of the milk-producing animal.

23. A method of selecting a milk-producing animal for breeding, comprising:(a) performing the method of predicting the productivity of a milk-producing animal of any one of claims 1 to 15 and 22, and selecting the milk-producing animal or parent for breeding if the method predicts high productivity; or(b) performing the method of using a trained machine learning algorithm model to characterise a milk-producing animal as being of high productivity or low productivity of any one of claims 16 and 18 to 22, and selecting the milk-producing animal or parent for breeding if the method characterises the milk-producing animal as being of high productivity.

24. A computer program comprising instructions that, when executed by a computer system, instruct the computer system to perform the method of any one of claims 16 to 22.

25. A data processing apparatus comprising a processor configured to perform the method of any one of claims 16 to 22.

Images