Genomic prediction using machine learning: a comparison of the performance of regularized regression, ensemble, instance-based and deep learning methods on synthetic and empirical data.

BACKGROUND: The accurate prediction of genomic breeding values is central to genomic selection in both plant and animal breeding studies. Genomic prediction involves the use of thousands of molecular markers spanning the entire genome and therefore requires methods able to efficiently handle high dimensional data. Not surprisingly, machine learning methods are becoming widely advocated for and used in genomic prediction studies. These methods encompass different groups of supervised and unsupervised learning methods. Although several studies have compared the predictive performances of individual methods, studies comparing the predictive performance of different groups of methods are rare. However, such studies are crucial for identifying (i) groups of methods with superior genomic predictive performance and assessing (ii) the merits and demerits of such groups of methods relative to each other and to the established classical methods. Here, we comparatively evaluate the genomic predictive performance and informally assess the computational cost of several groups of supervised machine learning methods, specifically, regularized regression methods, deep, ensemble and instance-based learning algorithms, using one simulated animal breeding dataset and three empirical maize breeding datasets obtained from a commercial breeding program. RESULTS: Our results show that the relative predictive performance and computational expense of the groups of machine learning methods depend upon both the data and target traits and that for classical regularized methods, increasing model complexity can incur huge computational costs but does not necessarily always improve predictive accuracy. Thus, despite their greater complexity and computational burden, neither the adaptive nor the group regularized methods clearly improved upon the results of their simple regularized counterparts. This rules out selection of one procedure among machine learning methods for routine use in genomic prediction. The results also show that, because of their competitive predictive performance, computational efficiency, simplicity and therefore relatively few tuning parameters, the classical linear mixed model and regularized regression methods are likely to remain strong contenders for genomic prediction. CONCLUSIONS: The dependence of predictive performance and computational burden on target datasets and traits call for increasing investments in enhancing the computational efficiency of machine learning algorithms and computing resources.

Relationship between Physicochemical and Cooking Quality Parameters with Estimated Glycaemic Index of Rice Varieties.

Rice is a significant staple food in the basic diet of the global population that is considered to have a high glycaemic index. The study of the physical and chemical parameters in rice that are related to the starch digestion process, which allows us to quickly predict the glycaemic index of varieties, is a major challenge, particularly in the classification and selection process. In this context, and with the goal of establishing a relationship between physicochemical properties and starch digestibility rates, thus shedding light on the connections between quality indicators and their glycaemic impact, we evaluated various commercial rice types based on their basic chemical composition, physicochemical properties, cooking parameters, and the correlations with digestibility rates. The resistant starch, the gelatinization temperature and the retrogradation (setback) emerge as potent predictors of rice starch digestibility and estimated glycaemic index, exhibiting robust correlations of r = -0.90, r = -0.90, and r = -0.70 (p ≤ 0.05), respectively. Among the rice types, Long B and Basmati stand out with the lowest estimated glycaemic index values (68.44 and 68.10), elevated levels of resistant starch, gelatinization temperature, and setback values. Furthermore, the Long B showcases the highest amylose, while the Basmati with intermediate, revealing intriguingly strong grain integrity during cooking, setting it apart from other rice varieties.

Nutritional Properties of Baobab Pulp from Different Angolan Origins.

The baobab tree (Adansonia digitata L.) is found widely in the forests and savannas of sub-Saharan Africa. The baobab fruit has a sour and slightly sweet taste and is widely consumed by the natives, thus containing a high nutritional value and providing a source of income for rural people. This study aimed to compare the nutritional composition of baobab fruit pulp from different localities in the Namibe province (Angola). Twenty samples of baobab pulp were collected in markets of the four municipalities of Namibe. The results obtained showed that there is some geographic location dependence on nutritional and functional composition. The municipality of Camucuio showed samples with higher fibre content (56.62 g/100 g) and vitamin C (288.9 mg/100 g). Samples from the Virei municipality stood out for their antioxidant activity (1936 mmol TE/100 g), high K content (42.4 mg/g) and higher values of protein (2.42 g/100 g). The samples collected in the municipality of Bibala stood out for their high contents of carbohydrates (28.1 g/100 g), total phenolic compounds (972 mg GAE/100 g) and Ca (3.80 mg/g). Despite the differences in origin, the high nutritional value of baobab fruit has the potential to improve the diet of thousands of people in Africa qualitatively.

Rice Compounds with Impact on Diabetes Control.

Rice is one of the most cultivated and consumed cereals worldwide. It is composed of starch, which is an important source of diet energy, hypoallergenic proteins, and other bioactive compounds with known nutritional functionalities. Noteworthy is that the rice bran (outer layer of rice grains), a side-stream product of the rice milling process, has a higher content of bioactive compounds than white rice (polished rice grains). Bran functional ingredients such as γ-oryzanol, phytic acid, ferulic acid, γ-aminobutyric acid, tocopherols, and tocotrienols (vitamin E) have been linked to several health benefits. In this study, we reviewed the effects of rice glycemic index, macronutrients, and bioactive compounds on the pathological mechanisms associated with diabetes, identifying the rice compounds potentially exerting protective activities towards disease control. The effects of starch, proteins, and bran bioactive compounds for diabetic control were reviewed and provide important insights about the nutritional quality of rice-based foods.

A robust AMMI model for the analysis of genotype-by-environment data.

MOTIVATION: One of the most widely used models to analyse genotype-by-environment data is the additive main effects and multiplicative interaction (AMMI) model. Genotype-by-environment data resulting from multi-location trials are usually organized in two-way tables with genotypes in the rows and environments (location-year combinations) in the columns. The AMMI model applies singular value decomposition (SVD) to the residuals of a specific linear model, to decompose the genotype-by-environment interaction (GEI) into a sum of multiplicative terms. However, SVD, being a least squares method, is highly sensitive to contamination and the presence of even a single outlier, if extreme, may draw the leading principal component towards itself resulting in possible misinterpretations and in turn lead to bad practical decisions. Since, as in many other real-life studies the distribution of these data is usually not normal due to the presence of outlying observations, either resulting from measurement errors or sometimes from individual intrinsic characteristics, robust SVD methods have been suggested to help overcome this handicap. RESULTS: We propose a robust generalization of the AMMI model (the R-AMMI model) that overcomes the fragility of its classical version when the data are contaminated. Here, robust statistical methods replace the classic ones to model, structure and analyse GEI. The performance of the robust extensions of the AMMI model is assessed through a Monte Carlo simulation study where several contamination schemes are considered. Applications to two real plant datasets are also presented to illustrate the benefits of the proposed methodology, which can be broadened to both animal and human genetics studies. AVAILABILITY AND IMPLEMENTATION: Source code implemented in R is available in the supplementary material under the function r-AMMI. CONTACT: paulocanas@gmail.com SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.