A gene expression estimator of intramuscular fat percentage for use in both cattle and sheep
© Guo et al.; licensee BioMed Central Ltd. 2014
Received: 19 December 2013
Accepted: 12 June 2014
Published: 16 June 2014
The expression of genes encoding proteins involved in triacyglyceride and fatty acid synthesis and storage in cattle muscle are correlated with intramuscular fat (IMF)%. Are the same genes also correlated with IMF% in sheep muscle, and can the same set of genes be used to estimate IMF% in both species?
The correlation between gene expression (microarray) and IMF% in the longissimus muscle (LM) of twenty sheep was calculated. An integrated analysis of this dataset with an equivalent cattle correlation dataset and a cattle differential expression dataset was undertaken. A total of 30 genes were identified to be strongly correlated with IMF% in both cattle and sheep. The overlap of genes was highly significant, 8 of the 13 genes in the TAG gene set and 8 of the 13 genes in the FA gene set were in the top 100 and 500 genes respectively most correlated with IMF% in sheep, P-value = 0. Of the 30 genes, CIDEA, THRSP, ACSM1, DGAT2 and FABP4 had the highest average rank in both species. Using the data from two small groups of Brahman cattle (control and Hormone growth promotant-treated [known to decrease IMF% in muscle]) and 22 animals in total, the utility of a direct measure and different estimators of IMF% (ultrasound and gene expression) to differentiate between the two groups were examined. Directly measured IMF% and IMF% estimated from ultrasound scanning could not discriminate between the two groups. However, using gene expression to estimate IMF% discriminated between the two groups. Increasing the number of genes used to estimate IMF% from one to five significantly increased the discrimination power; but increasing the number of genes to 15 resulted in little further improvement.
We have demonstrated the utility of a comparative approach to identify robust estimators of IMF% in the LM in cattle and sheep. We have also demonstrated a number of approaches (potentially applicable to much smaller groups of animals than conventional methods) to using gene expression to rank animals for IMF% within a single farm/treatment, or to estimate differences in IMF% between two farms/treatments.
Consumers are prepared to pay more for meat with superior eating qualities . Intramuscular fat (IMF), the flecks and streaks of fat within the lean sections of meat, which is also known as marbling, is associated with juiciness and flavour . Recent research has shown that increased IMF% could dramatically improve the tenderness of lamb carcasses 5 days post-slaughter . But compared to beef-related research (see [4, 5]), few publications have focussed on the molecular mechanism of IMF deposition in sheep. In the past few years, only FABP3 (H-FABP), PPARG, DGAT1, LPL, ACACA, FASN (FAS), FABP4, CPT1B and SCD have been reported to directly influence IMF% status in sheep LM [6–9]. Thus, based on the limited information from sheep, it is hard to identify a set of genes to estimate IMF%.
In our previous studies in cattle, three gene sets, designated as the “TAG gene set” (triglyceride synthesis and storage), the “FA gene set” (fatty acid synthesis and storage) and the “PPARG gene set” (Peroxisome proliferator-activated receptor gamma), were identified based on the expression profiles of the genes in the LM across development in two crosses . The expression of genes from these three gene sets, in particular the TAG gene set, was correlated with IMF deposition in cattle LM . The TAG gene set was used to identify the effect of HGP (hormone growth promotant) treatment, site (New South Wales [NSW] and Western Australia [WA]) and Calpain/calpastatin genotype on IMF% .
Cattle and sheep are evolutionarily closely related  and are expected to exhibit many common physiological characteristics. In this study, we hypothesised that the genes in the TAG, FA and PPARG gene sets identified in cattle could also be applied to estimate IMF% in sheep. Furthermore, based on these gene sets, we evaluated the utility of single and small sets of genes to estimate IMF% in small groups of animals from both species.
Materials and methods
Use of animals and the proce stry & Investment New South Wales (NSW) Orange Agriculture Institute Animal Ethics Committee, Commonwealth Scientific and Industrial Organisation (CSIRO) Rockhampton Animal Experimentation Ethics Committee, and the Department of Agriculture and Food, Western Australia (WA) Animal Ethics Committee.
Sheep correlation dataset
The design of the experiment has been described previously . Briefly, 20 sheep were randomly assigned to five groups, four groups of treated animals received implants containing a combination of ~42 mg trenbolone acetate (TBA) and ~4.2 mg 17-βestradiol (E2), or ~50 mg TBA alone, or ~10 mg E2 alone at the start of the trial or 20 mg oxytocin delivered by Alzet osmotic pump over 30 d at the start of the experiment and again after 30 and 60 d. Following slaughter, 50 mg of LM tissue (between the 12th and 13th rib) and the strip loins (6th to 9th rib) were collected from the right sides of the carcasses for RNA preparation and meat quality analyses, respectively. IMF% was measured in duplicate on each sample by gas chromatography (GC) as previously described . Gene expression was measured using the Bovine Oligo Microarray Chip (Bovine 4x44K) from Agilent Technologies (Santa Clara CA, USA and will be described in detail elsewhere [Kongsuwan et al., in preparation]). The same platform was used for the two bovine gene expression datasets described below. The Bovine Oligo Microarray platform was used as it has a larger coverage of genes than the equivalent sheep array and using the same platform simplifies data integration and analysis.
Cattle correlation dataset
Correlation between gene expression and IMF% in the LM muscle in a group of 48 intensively fed Brahman steers, including three tenderness genotypes, an environment contrast (growth at two different sites, New South Wales [NSW] and Western Australia [WA]) and with and without a hormone growth promotant (HGP) treatment, has been described previously . IMF% was measured by Near Infrared Spectrophotometry (NIRS), duplicate measurements on single samples, as previously described . Ultrasound estimation of IMF% was undertaken as previously described , values were the mean of five measurements.
Cattle DE dataset
Differential expression (DE) of genes from two cattle crosses with high and medium marbling, Wagyu cross Hereford and Piedmontese cross Hereford respectively, has been described previously . The cattle in this dataset were sampled at 25 mo of age whilst at pasture.
Statistics and bioinformatics
The correlation between gene expression and IMF% was calculated using the “CORREL” function in Microsoft Excel. Student’s t-test of significance was calculated using the “TTEST” function (one tailed) in Microsoft Excel.
P-values for the hypergeometric distribution for a specified number of successes in a population sample were calculated using the Excel “HYPERGEOMDIST” function.
Gene enrichment analysis was undertaken by using GOrilla network tools which uses a hypergeometric statistic to quantify functional enrichment in ranked gene lists . P-values, and the false discovery rate (FDR) Q-values calculated using the Benjamini and Hochberg method , were provided in the results output of the GOrilla website .
Cluster analysis was undertaken using an expectation-maximization mixture analysis algorithm (EMMIX) . All three datasets were linearly rescaled to a mean of zero and a range from −0.5 to 0.5 before analysis.
The z-score normalization was used to minimise the impact of differences in levels of expression and dynamic range of expression of genes from the combed gene expression data, individual gene expression values (log2) were normalised by dividing its difference from the mean of each measurement (across the whole set or subsets of the animals) with the relevant standard deviation using Microsoft Excel.
The random sampling and calculation of mean correlation was carried out in MATLAB software R2012a using custom scripts. Random controls with 5 genes were sampled 100,000 times by using random sampling from the rescaled cattle correlation dataset, this process was repeated 10 times. Those possessing higher correlation with IMF% in cattle in each sampling process were investigated. Then correlation of average gene expression with IMF% in these random controls in sheep was calculated.
The significance of gene rankings between groups was calculated using the Mann–Whitney Test web tool .
Sample size determination was then performed to estimate the minimum number of animals required to significantly differentiate two groups at a P-value of 0.05 and confidence interval of 95% , A Microsoft Excel spreadsheet “LaMorte’s Power Calculator” downloaded from the web site  was used.
Results and discussion
Expression of genes in the TAG, FA and PPARG gene sets was correlated with IMF% in sheep
Rankings of genes in various datasets and average ranking
cell death-inducing dffa-like effector a
acyl-CoA synthetase medium-chain family member 1
fatty acid binding protein 4,adipocyte
tumor suppressor candidate 5
mal, T-cell differentiation protein 2
peroxisome proliferator-activated receptor gamma
diacylglycerol o-acyltransferase 2
phosphate O-acyltransferase 2
fatty acid synthase
thyroid hormone responsive
ELOVL fatty acid elongase 6
cell death-inducing DFFA-like effector c
cytochrome b5 type A (microsomal)
betaine-homocysteine S-methyltransferase 2
retinol binding protein 4, plasma
acyl-CoA synthetase short-chain family member 2
arylsulfatase family, member K
acetyl-CoA carboxylase alpha
atp citrate lyase
carnitine palmitoyltransferase 2
phosphoenolpyruvate carboxykinase 1
protein tyrosine phosphatase-like, member b
phosphoenolpyruvate carboxykinase 2 (mitochondrial)
integrator complex subunit 9
phosphodiesterase 3B, cGMP-inhibited
alkaline ceramidase 3
isocitrate dehydrogenase 1 (NADP+), soluble
glutathione S-transferase alpha 1
sulfotransferase family, cytosolic, 1A, phenol-preferring, member 1
alanyl (membrane) aminopeptidase
hydroxysteroid (17-beta) dehydrogenase 12
3-hydroxybutyrate dehydrogenase, type 1
acyl-CoA synthetase short-chain family member 3
malic enzyme 1, NADP(+)-dependent, cytosolic
S100 calcium binding protein G
insulin induced gene 1
glycerol-3-phosphate acyltransferase, (mitochondrial)
fructose-1, 6-bisphosphatase 1
CCAAT/enhancer binding protein (C/EBP), alpha
The sheep correlation dataset was generated from a small group of animals, therefore it is unavoidably noisy. Hence, integrated analysis of the cattle and sheep datasets was undertaken in order to identify robust genes for use in both species.
A cluster analysis of all the genes based on their correlation coefficients in cattle and sheep and DE values in cattle was undertaken using EMMIX . Each gene was assigned a clustering parameter ranging from 0 to 1, the probability for location in the two alternative clusters. The smaller group was defined as Cluster A, the other as Cluster B (Additional file 1: Table S1). Genes with positive values in all three datasets in both clusters were selected, and then submitted to GOrilla for gene ontology enrichment analysis.
Gene enrichment analysis of EMMIX Cluster A
Ninteen genes from the previously described TAG, FA and PPARG gene sets from cattle were included in the 30 genes. The remaining 11 genes included two categories of genes. The first category included genes with a well characterised and important role in lipid metabolic process and with relatively high correlation/DE coefficients, such as LPL and G0S2[21, 22]. The second category contained genes with very low correlation/DE coefficients (close to 0) and included in the lipid metabolic process GO term, such as ARSK, BDH1 and SULT1A1[23–25].
Of the genes reported in the literature to be important in sheep IMF deposition, PPARG, LPL, ACACA, FABP4, FASN (FAS) and SCD were included in the top 30, FABP3 (H-FABP), DGAT1, and CPT1B[4, 6–9] were not.
Correlation of CIDEA and IMF gene set(s) with IMF% in both cattle and sheep
CIDEA was the highest ranked gene based on correlation coefficients in both the cattle and sheep datasets (Table 1). Thus CIDEA is the best candidate for use as a single gene estimator of IMF% in both species. However a single gene may not be the best estimator of IMF%. Combinations of genes may be a better solution because they will provide multiple measurements of the lipid synthesis and storage pathway thereby reducing the measurement error. The set of 30 genes described above was used to identify a gene combination(s) able to estimate IMF% equally well in both cattle and sheep. CIDEA, THRSP, ACSM1, DGAT2 and FABP4 from the TAG gene set were selected based on their combined ranking in both the cattle and sheep correlation datasets (Table 2). Consistent with our findings, two of these five genes (DGAT2 and FABP4) have also been identified to be correlated with IMF in cattle previously [26, 27]. We tested the top 2, top 3, top 4 and top 5 genes above (defined as the “IMF 2–5 gene sets”) to determine whether combinations of the top genes had a higher correlation with IMF% in both species than CIDEA alone. A simple model (average of rescaled gene expression values) was used to combine the data from the different genes.
Correlation between gene expression and IMF%
IMF 5 gene set
NSW control group
As previously described , the 48 Brahman cattle were divided into four subgroups by experimental site (NSW and WA) and treatment (with and without HGP). In the NSW control subgroup, consistent with previous analysis , no significantly positive correlation between expression of any of the 5 top ranking genes and IMF% was observed (Table 3 and data not shown). In addition, in this group of animals, none of the “IMF 2–5 gene sets” showed significant correlation with IMF% (Table 3 and data not shown). This is probably due to an environmental factor such as disease or nutrition. For this reason, we repeated the analysis using the remaining three subgroups of cattle (NSW HGP-treated, WA control and WA HGP-treated), 36 animals in total. Expression of CIDEA was now as correlated with IMF% (0.61) as the “IMF 2–5 gene sets” (0.60-0.62) when the NSW control group was not included (Table 3).
However, the group of cattle excluding the NSW control subgroup, and the sheep may not be representative of the spectrum of animals in real production systems. Thus, we repeated the analysis on all the 48 Brahman steers and 20 sheep. In the cattle, the expression of the IMF 5 gene set showed a slightly higher correlation with IMF% than CIDEA alone (Table 3). In sheep, the IMF 5 gene set showed similarly high correlation with IMF% as CIDEA alone (Table 3).
Overall, on the basis of the correlation of gene expression with IMF%, CIDEA alone and the IMF 5 gene set performed very similarly in sheep and in cattle and in the subsets of cattle.
Relationship between gene expression and IMF% in both cattle and sheep
Is there a simple relationship between CIDEA and IMF 5 gene set gene expression and IMF% in both species facilitating the development of a gene-based test for sorting animals by their estimated IMF% in the LM?
Applications of gene estimator(s) of IMF% in both cattle and sheep
Comparison of the performance of different measures and estimators of IMF%
Number of animals1
WA control subgroup
WA HGP subgroup
Predicted experiment size3
NIRS measured IMF%
2.37 ± 1.004
1.90 ± 0.83
Ultrasound estimated IMF%
2.66 ± 0.724
2.93 ± 0.54
NIRS measured IMF%
2.07 ± 0.774
1.79 ± 0.54
IMF% calculated by CIDEA formula
2.30 ± 1.196
1.60 ± 1.14
IMF% calculated by IMF 5 gene set formula
2.66 ± 1.046
1.30 ± 0.92
Ranking animals using CIDEA
9.50 ± 6.627
13.20 ± 6.16
Ranking animals using IMF 5 gene set
7.20 ± 5.478
15.10 ± 4.99
Ranking animals using TAG gene set
7.00 ± 4.528
15.30 ± 5.48
13.09 ± 0.447
12.83 ± 0.38
IMF 5 gene set DE
0.25 ± 0.389
−0.27 ± 0.31
TAG gene set DE
0.28 ± 0.369
−0.27 ± 0.33
To demonstrate the two applications of our findings, the regression equations based on the relationship between IMF% and expression of CIDEA, or the IMF 5 gene set, were used to estimate the IMF% for each animal. For both approaches there was a significant difference between the means of the estimated IMF% of the two groups (P < 0.05), unlike for the NIRS measured IMF% on the same 22 animals (P = 0.34) (Table 4). Although the correlation between CIDEA expression and IMF%, and the expression of the IMF 5 gene set with IMF% were very similar (Table 3). The use of more genes appears to significantly improve the accuracy of the estimation of IMF% (Table 4).
Rather than calculating an estimated IMF%, animals can be ranked on the basis of the relative gene expression values. The results showed that there was a significant difference between the average rankings of animals in the WA HGP-treated and control groups using the IMF 5 gene set (P-value = 0.0026) and TAG gene set (P-value = 0.0017), but not using CIDEA alone (P-value = 0.1) (Table 4). Again the apparent accuracy of the ranking method was improved significantly by the inclusion of additional genes, although increasing to 15 genes provided little additional improvement.
To compare IMF% of animals between different farms, besides the approaches above, we could also compare the DE of CIDEA, IMF gene set or TAG gene sets between two groups of animals. Rescaled gene expression data was used for the DE calculation of the multiple genes in the IMF 5 gene set and TAG gene set. As in the ranking method above, increasing the number of genes from one to five increased the discrimination between the two groups, but increasing the number of genes to 15 had little further effect (Table 4).
Generally speaking, in both research and production settings, the number of animals tested is the major determinant of the cost of collecting phenotypes. Methods which reduce the number of animals required to be tested and/or which can be conducted without sacrificing the animal will reduce the costs of phenotyping substantially. Using the approach of LaMorte , we estimated the sample size required to detect an affect at P < 0.05 and a confidence interval of >95% for all of the analysis methods (Table 4). Given the small size of the datasets the sample sizes for approaches using gene expression are likely to be overestimated. No reliable estimate of the sample size could be made for the use of ultrasound as the available data confirms that ultrasound is inaccurate for animals with low IMF%. To detect the effect using NIRS measured IMF% a large number of animals are required (Table 4). Using gene expression of CIDEA a slightly smaller sample size may be adequate. However, the use of five genes substantially reduced the predicted sample size, suggesting that around one eighth of the number of animals may be required to detect the effect of HGPs on IMF%. This improvement in performance and hence reduction in experiment size may be because the use of a multiple gene set effectively provides multiple measurements of the phenotype (deposition of TAG in lipid droplets in intramuscular adipocytes) leading to a reduced measurement error than using CIDEA alone, or the mean of duplicate NIRS measurements.
By integrating data from cattle and sheep we have identified a set of 30 genes with robust correlation with IMF% in both cattle and sheep LM. Based on this gene set, we identified CIDEA as the gene whose expression was most correlated with IMF% in both cattle and sheep. Whilst CIDEA alone could be used to estimate IMF%, it is of similar utility to NIRS measured IMF%. In contrast, the non-invasive technique of ultrasound did not perform adequately on animals with low IMF%. By combining the data from 5 genes apparently improved estimates of IMF% could be calculated, with a commensurate reduction in the experiment size required to detect the impact of a treatment on IMF%. The five gene set can be used to estimate IMF% (based on the proposed relationship between the expression of the IMF 5 gene set, IMF deposition rate and IMF%) from biopsy as well as post slaughter samples, and on samples from animals with low IMF%, such as the Brahmans used in this work and younger animals of higher marbling breeds. The approach to phenotyping animals using gene expression shows promise as an alternative to current approaches for the measurement/estimation of IMF% in both cattle and sheep.
In addition, we have described a potentially generic approach to the development of robust gene expression phenotypes for other phenotypes of industry importance. The pipeline is as follows: calculate the correlation between gene expression and a phenotype (IMF% in this paper) and/or DE in two or more different groups of animals with significantly different experimental structures and phenotypic performance to generate the corresponding datasets. Then rank the genes based on the coefficients above in each dataset to primarily select a group of genes highly correlated with this phenotype and with each other across the different groups of animals. Lastly, optimise this group of genes based on their biological function to identify a gene set with appropriate size.
This project was partially supported by the CRC for Beef Genetic Technologies. The authors would like to thank Dr Nicholas Hudson for discussions, Dr Antonio Reverter for his help with EMMIX calculations, Mr. Zhenliang Ma at The Faculty of Engineering, Architecture & Information Technology (EAIT) of The University of Queensland Australia for his advice on programming, Dr Yutao Li for help with statistical techniques and Matthew Wolcott for access to unpublished ultrasound data.
- Pethick DW, Pleasants AB, Gee AM, Hopkins DL, Ross IR: Eating quality of commercial meat cuts from Australian lambs and sheep. Proc New Zeal Soc An. 2006, 66: 363-367.Google Scholar
- Hocquette J, Gondret F, Baéza E, Médale F, Jurie C, Pethick D: Intramuscular fat content in meat-producing animals: development, genetic and nutritional control, and identification of putative markers. Animal. 2010, 4: 303-319.View ArticlePubMedGoogle Scholar
- Warner RD, Jacob RH, Edwards JEH, McDonagh M, Pearce K, Geesink G, Kearney G, Allingham P, Hopkins DL, Pethick DW: Quality of lamb meat from the Information Nucleus Flock. Anim Prod Sci. 2010, 50: 1123-1134.View ArticleGoogle Scholar
- De Jager N, Hudson NJ, Reverter A, Barnard R, Cafe LM, Greenwood PL, Dalrymple BP: Gene expression phenotypes for lipid metabolism and intramuscular fat in skeletal muscle of cattle. J Anim Sci. 2013, 91: 1112-1128.View ArticlePubMedGoogle Scholar
- Wang YH, Bower NI, Reverter A, Tan SH, De Jager N, Wang R, McWilliam SM, Cafe LM, Greenwood PL, Lehnert SA: Gene expression patterns during intramuscular fat development in cattle. J Anim Sci. 2009, 87: 119-130.View ArticlePubMedGoogle Scholar
- Huang ZG, Xiong L, Liu ZS, Qiao Y, Liu SR, Ren HX, Xie Z, Liu GQ, Li XB: The developmental changes and effect on IMF content of H-FABP and PPARgamma mRNA expression in sheep muscle. Yi Chuan Xue Bao. 2006, 33: 507-514.PubMedGoogle Scholar
- Qiao Y, Huang Z, Li Q, Liu Z, Hao C, Shi G, Dai R, Xie Z: Developmental changes of the FAS and HSL mRNA expression and their effects on the content of intramuscular fat in Kazak and Xinjiang sheep. J Genet Genomics. 2007, 34: 909-917.View ArticlePubMedGoogle Scholar
- Xu QL, Chen YL, Ma RX, Xue P: Polymorphism of DGAT1 associated with intramuscular fat-mediated tenderness in sheep. J Sci Food Agric. 2009, 89: 232-237.View ArticleGoogle Scholar
- Dervishi E, Serrano C, Joy M, Serrano M, Rodellar C, Calvo JH: The effect of feeding system in the expression of genes related with fat metabolism in semitendinous muscle in sheep. Meat Sci. 2011, 89: 91-97.View ArticlePubMedGoogle Scholar
- Kijas JW, Menzies M, Ingham A: Sequence diversity and rates of molecular evolution between cattle and sheep genes. Anim Genet. 2006, 37: 171-174.View ArticlePubMedGoogle Scholar
- Kongsuwan K, Knox MR, Allingham PG, Pearson R, Dalrymple BP: The effect of combination treatment with trenbolone acetate and estradiol-17beta on skeletal muscle expression and plasma concentrations of oxytocin in sheep. Domest Anim Endocrinol. 2012, 43: 67-73.View ArticlePubMedGoogle Scholar
- Cafe LM, McIntyre BL, Robinson DL, Geesink GH, Barendse W, Greenwood PL: Production and processing studies on calpain-system gene markers for tenderness in Brahman cattle: 1. Growth, efficiency, temperament, and carcass characteristics. J Anim Sci. 2010, 88: 3047-3058.View ArticlePubMedGoogle Scholar
- Upton W, Donoghue K, Graser H, Johnston D: Ultrasound proficiency testing. Proc Assoc Advmt Anim Breed Genet. 1999, 13: 341-344.Google Scholar
- Eden E, Navon R, Steinfeld I, Lipson D, Yakhini Z: GOrilla: a tool for discovery and visualization of enriched GO terms in ranked gene lists. BMC Bioinformatics. 2009, 10: 48.PubMed CentralView ArticlePubMedGoogle Scholar
- Benjamini Y, Hochberg Y: Controlling the False Discovery Rate - a Practical and Powerful Approach to Multiple Testing. J Roy Stat Soc B Met. 1995, 57: 289-300.Google Scholar
- GOrrila - a tool for identifying enriched GO terms. http://cbl-gorilla.cs.technion.ac.il/.
- McLachlan GJ, Peel D, Basford KE, Adams P: The EMMIX software for the fitting of mixtures of normal and t-components. J Stat Softw. 1999, 4: 1-14.View ArticleGoogle Scholar
- Mann–Whitney Test. http://www.vassarstats.net/utest.html.
- Dell RB, Holleran S, Ramakrishnan R: Sample size determination. Ilar J. 2002, 42: 207-213.View ArticleGoogle Scholar
- Lamorte Calculator. http://www.uml.edu/Images/LaMorte%20calculator_tcm18-37802.xls.
- Hung YP, Lee NY, Lin SH, Chang HC, Wu CJ, Chang CM, Chen PL, Lin HJ, Wu YH, Tsai PJ, Tsai YS, Ko WC: Effects of PPARgamma and RBP4 gene variants on metabolic syndrome in HIV-infected patients with anti-retroviral therapy. PLoS One. 2012, 7: e49102.PubMed CentralView ArticlePubMedGoogle Scholar
- Nielsen TS, Kampmann U, Nielsen RR, Jessen N, Orskov L, Pedersen SB, Jorgensen JO, Lund S, Moller N: Reduced mRNA and protein expression of perilipin A and G0/G1 switch gene 2 (G0S2) in human adipose tissue in poorly controlled type 2 diabetes. J Clin Endocrinol Metab. 2012, 97: E1348-1352.View ArticlePubMedGoogle Scholar
- Obaya AJ: Molecular cloning and initial characterization of three novel human sulfatases. Gene. 2006, 372: 110-117.View ArticlePubMedGoogle Scholar
- Martinez-Outschoorn UE, Lin Z, Whitaker-Menezes D, Howell A, Sotgia F, Lisanti MP: Ketone body utilization drives tumor growth and metastasis. Cell Cycle. 2012, 11: 3964-3971.PubMed CentralView ArticlePubMedGoogle Scholar
- McGill MR, Jaeschke H: Metabolism and Disposition of Acetaminophen: Recent Advances in Relation to Hepatotoxicity and Diagnosis. Pharm Res. 2013, 30 (9): 2174-2187.PubMed CentralView ArticlePubMedGoogle Scholar
- Jeong J, Kwon E, Im S, Seo K, Baik M: Expression of fat deposition and fat removal genes is associated with intramuscular fat content in longissimus dorsi muscle of Korean cattle steers. J Anim Sci. 2012, 90: 2044-2053.View ArticlePubMedGoogle Scholar
- da Costa AS, Pires VM, Fontes CM, Prates JAM: Expression of genes controlling fat deposition in two genetically diverse beef cattle breeds fed high or low silage diets. BMC Vet Res. 2013, 9: 118.PubMed CentralView ArticlePubMedGoogle Scholar
- Hunter R: Hormonal growth promotant use in the Australian beef industry. Anim Prod Sci. 2010, 50: 637-659.View ArticleGoogle Scholar
- Cafe LM, McIntyre BL, Robinson DL, Geesink GH, Barendse W, Pethick DW, Thompson JM, Greenwood PL: Production and processing studies on calpain-system gene markers for tenderness in Brahman cattle: 2. Objective meat quality. J Anim Sci. 2010, 88: 3059-3069.View ArticlePubMedGoogle Scholar
- Sample size calculation in research. http://www.uml.edu/docs/sample%20size%20calcs%20LaMorte_tcm18-37807.doc.
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly credited. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.