Cereals

The FFQ, originally developed for the TLGS, was a Willett-format questionnaire modified based on Iranian food items²⁵ and contains questions about average consumption and frequency for 168 food items during the past year.⁷ The food items were chosen according to the most frequently consumed items in the national food consumption survey in Iran.²⁵ Because different recipes are used for food preparation, the FFQ was based on food items rather than dishes, eg, beans, different meats and oils, and rice. Subjects indicated their food consumption frequencies on a daily basis (eg, for bread), weekly basis (eg, for rice and meat), monthly basis (eg, for fish), yearly basis (eg, for organ meats), or a never/seldom basis according to portion sizes that were provided in the FFQ. For each food item on the FFQ, a portion size was specified using USDA serving sizes (eg, bread, 1 slice; apple, 1 medium; dairy, 1 cup) whenever possible; if this was not possible, household measures (eg, beans, 1 tablespoon; chicken meat, 1 leg, breast, or wing; rice, 1 large, medium, or small plate) were chosen. Table 1shows food items and portion sizes used in the FFQ. Trained dietary interviewers with at least 3 of experience in the Nationwide Food Consumption Survey project²⁵ or TLGS^{26 (link)} administered the FFQs and 24-hour DRs during face-to-face interviews. The interviewer read out the food items on the FFQ, and recorded their serving size and frequency. The interview session took about 45 minutes. The interviewer for FFQ1 and FFQ2 was the same for each participant. Daily intakes of each food item were determined based on the consumption frequency multiplied by the portion size or household measure for each food item.²⁷ The weight of seasonal foods, like some fruits, was estimated according to the number of seasons when each food was available.
Dietary data were also collected monthly by means of twelve 24-hour DRs that lasted for 20 minutes on average. For all subjects, 2 formal weekend day (Thursday and Friday in Iran) and 10 weekdays were recalled. All recall interviews were performed at subjects’ homes to better estimate the commonly used household measures and to limit the number of missing subjects. Detailed information about food preparation methods and recipe ingredients were considered by interviewers. To prevent subjects from intentionally altering their regular diets, participants were informed of the recall meetings with dietitians during the evening before the interview. All recalls were checked by investigators, and ambiguities were resolved with the subjects. Mixed dishes in 24-hour DRs were converted into their ingredients according to the subjects’ report on the amount of the food item consumed, thus taking into account variations in meal preparation recipes. For instance, broth or soup ingredients—usually vegetables (carrot or green beans), noodles, barley, etc.—differed according to subjects’ meal preparation. Because the only available Iranian food composition table (FCT)²⁸ analyzes a very limited number of raw food items and nutrients, we used the USDA FCT²⁹ as the main FCT; the Iranian FCT was used as an alternative for traditional Iranian food items, like kashk, which are not included in the USDA FCT.
The food items on the FFQ and DR were grouped according to their nutrient contents, based on other studies,^{30 (link)} and modified according to our dietary patterns. Seventeen food groups were thus obtained, as follows: 1) whole grains, 2) refined grains, 3) potatoes, 4) dairy products, 5) vegetables, 6) fruits, 7) legumes, 8) meats, 9) nuts and seeds, 10) solid fat, 11) liquid oil, 12) tea and coffee, 13) salty snacks, 14) simple sugars, 15) honey and jams, 16) soft drinks, and 17) desserts and snacks (Table 1). The 168 food items on the FFQ were allocated to these 17 food groups, and the amounts in grams of each item were summed to obtain the daily intake of each food group.

We created an overall plant-based diet index (PDI), a healthful plant-based diet index (hPDI), and an unhealthful plant-based diet index (uPDI). The procedure we used to create these indices is similar to the one used by Martínez-González et al. [13 (link)]; their “provegetarian food pattern” is similar in composition to our PDI. Frequencies of consumption of each food were converted into servings consumed per day. Then the number of servings of foods that belonged to each of 18 food groups were added up. The 18 food groups were created on the basis of nutrient and culinary similarities, within larger categories of animal foods and healthy and less healthy plant foods. We distinguished between healthy and less healthy plant foods using existing knowledge of associations of the foods with T2D, other outcomes (CVD, certain cancers), and intermediate conditions (obesity, hypertension, lipids, inflammation). Plant foods not clearly associated in one direction with several health outcomes, specifically alcoholic beverages, were not included in the indices. We also excluded margarine from the indices, as its fatty acid composition has changed over time from high trans fat to high unsaturated fat. We controlled for alcoholic beverages and margarine consumption in the analysis.
Healthy plant food groups included whole grains, fruits, vegetables, nuts, legumes, vegetable oils, and tea/coffee, whereas less healthy plant food groups included fruit juices, sugar-sweetened beverages, refined grains, potatoes, and sweets/desserts. Animal food groups included animal fats, dairy, eggs, fish/seafood, meat (poultry and red meat), and miscellaneous animal-based foods.
S1 Table details examples of foods constituting the food groups. The 18 food groups were divided into quintiles of consumption, and each quintile was assigned a score between 1 and 5. For PDI, participants received a score of 5 for each plant food group for which they were above the highest quintile of consumption, a score of 4 for each plant food group for which they were above the second highest quintile but below the highest quintile, and so on, with a score of 1 for consumption below the lowest quintile (positive scores). On the other hand, participants received a score of 1 for each animal food group for which they were above the highest quintile of consumption, a score of 2 for each animal food group for which they were between the highest and second highest quintiles, and so on, with a score of 5 for consumption below the lowest quintile (reverse scores). For hPDI, positive scores were given to healthy plant food groups, and reverse scores to less healthy plant food groups and animal food groups. Finally, for uPDI, positive scores were given to less healthy plant food groups, and reverse scores to healthy plant food groups and animal food groups. The 18 food group scores for an individual were summed to obtain the indices, with a theoretical range of 18 (lowest possible score) to 90 (highest possible score). The observed ranges at baseline were 24–85 (PDI), 28–86 (hPDI), and 27–90 (uPDI) across the cohorts. The indices were analyzed as deciles, with energy intake adjusted at the analysis stage.

We performed genomic breeding value estimation (GEBV) and hybrid prediction with wheat data, and the results were compared to other genomic selection and mixed model software, including rrBLUP [13 ], ASReml [21 ], regress (used by synbreed as well) [17 ,18 (link)], EMMREML [19 ], MCMCglmm [15 (link)], and BGLR [16 (link)]. We used the wheat data contained in the R package BGLR consisting of 599 inbred lines genotyped with 1279 diversity array technology (DArT) markers [16 (link)]. Phenotypic data consisted of grain yield (GY) for the 599 lines from the historical CIMMYT's Global Wheat Program evaluated in four mega-environments.
From the 599 wheat lines, 179,101 distinct single crosses can be performed. Kinship-based BLUP prediction for the 599 lines were obtained using rrBLUP (ridge regression), ASReml (average information), regress (Newton-Raphson), EMMREML (modified EMMA), BGLR (using the Reproducing kernel Hilbert space [RKHS] kernel), MCMCglmm (Gibbs sampling) and the three algorithms implemented in sommer (AI, EM, and EMMA). Similarity among BLUPs using all software was performed in R and displayed in tables and figures [26 ]. The genomic estimated breeding values (GEBV) for each of the 599 inbred lines was used to predict the performance of possible crosses as the average among the breeding value of the parental lines. The mixed model fitted has the form:
$y=\mathit{X}\beta +\mathit{Z}u+\epsilon$
with variance:
$V\left(y\right)=V\left(\mathit{Z}u+\epsilon \right)={\mathbf{Z}\mathbf{G}\mathbf{Z}}^{\mathrm{\prime }}+\mathit{R}$
and the mixed model equations for this model are:
${\left[\begin{array}{cc}\mathit{X}\mathrm{\prime }{\mathit{R}}^{-1}\mathit{X}& \mathit{X}\mathrm{\prime }{\mathit{R}}^{-1}\mathit{Z}\\ \mathit{Z}\mathrm{\prime }{\mathit{R}}^{-1}\mathit{X}& {\mathit{Z}}^{\mathrm{\prime }}{\mathit{R}}^{-1}\mathit{Z}+{\mathit{G}}^{-1}\end{array}\right]}^{-1}\left[\begin{array}{c}\mathit{X}\prime {\mathit{R}}^{-1}y\\ \mathit{Z}\prime {\mathit{R}}^{-1}y\end{array}\right]=\left[\begin{array}{c}\beta \\ u\end{array}\right]$
Here, G = Kσ²_u, is the variance covariance matrix of the random effect u, from a multivariate normal distribution u ~ MVN(0, Kσ²_u), K being, in the genomics context, the additive or genomic relationship matrix (A or A_g). X and Z are incidence matrices for fixed and random effects respectively, and R is the matrix for residuals (here Iσ²_e). A mixed model with a single variance component other than the error (σ²_e) can be used to estimate the genetic variance (σ²_u) along with genotype BLUPs to exploit the genetic relationships between individuals coded in K (A). The genomic relationship matrix was constructed according to VanRaden where K = ZZ’/2Σp_i(1-p_i) [27 (link)]. Genotype BLUPs were calculated and considered equal to the GEBV and these were used to predict the performance of the 179,101 possible crosses as the average of parental genomic breeding values. We fitted this model using the sommer package by specifying the incidence and variance-covariance matrices and using the three algorithms implemented (AI, EM, EMMA). In addition, a five-fold cross validation was performed to calculate the predictive correlation for grain yield in the four mega environments available for the wheat data using the sommer package. In addition, heritability was estimated as h² = σ²_u / σ²_u + σ²_e.