Protein Quality - Amino Acid Score Validation
Donald Layman PhD, Emily Jennings-Dobbs MA, Shavawn Forester PhD RDN
2022-09-16
Special thanks to: Shazia Sathar MS and Olivia Comin
1 Introduction
Protein quality scores are important for choosing foods, creating diets and analyzing trends.The most commonly used scores are PDCAAS and DIAAS. Both of these scores, although well-established, are not not scalable nor feasible for broad-based use. In addition, they are based on flawed logic; a limiting amino acid based model. What is needed is an amino acid-based scoring system, a one to one design that reflects the ability of a food or meal to meet requirements. Amino acids, just like all other vitamins and minerals, are unique compounds on their own and should be treated as such. The proposed amino acid quality score gives equal waiting and importance to each essential amino acid. An amino acid-based quality score is possible and will have a profound positive impact on science and consumers alike.
PDCAAS is currently the most widely used protein quality score, it is calculated as the product of an amino acid score and protein digestibility score. We will be using the amino acid score from the PDCAAS calculation to directly and accurately validate our own scores.
2 Amino Acid Score Calculations
All EAA-3 and EAA-9 frameworks are based on the Institute of Medicine Recommended Dietary Allowance (RDA), please note that RDAs represent the minimum amount needed to “prevent depletion of the nutrient from the body” and may not necessarily indicate the optimum intake.
2.1 EAA-3 Calculation
We will be looking at four different calculations of the EAA-3 score:
- The Capped Average EAA-3 score is the average percent of the Recommended Dietary Allowance (RDA) of leucine, lysine, and methionine where the maximum percent of the RDA that can be met is 100%
- The Uncapped Average EAA-3 score represents the average percent of the Recommended Dietary Allowance (RDA) of leucine, lysine, and methionine
- The Capped minimum EAA-3 score is the minimum percentage of the RDA met by leucine, lysine, or methionine; capped at 100%
- The Uncapped minimum EAA-3 score is the minimum percentage of the RDA met by leucine, lysine, or methionine
Capped Average EAA-3:
\(\text{EAA-3}=\frac{(Min(\frac{\text{Leu Present}}{\text{Leu RDA}},1)+Min(\frac{\text{Lys Present}}{\text{Lys RDA}},1)+Min(\frac{\text{Met Present}}{\text{Met RDA}},1))}{3}\)
Uncapped Average EAA-3:
\(EAA-3 = \frac{(\frac{\text{Leu Present}}{\text{Leu RDA}}+\frac{\text{Lys Present}}{\text{Lys RDA}}+\frac{\text{Met Present}}{\text{Met RDA}})}{3}\)
Capped Minimum EAA-3:
\(\text{EAA-3}= Min(Min(\frac{\text{Leu Present}}{\text{Leu RDA}},1),Min(\frac{\text{Lys Present}}{\text{Lys RDA}},1),Min(\frac{\text{Met Present}}{\text{Met RDA}},1))\)
Uncapped Minimum EAA-3:
\(\text{EAA-3}= Min(\frac{\text{Leu Present}}{\text{Leu RDA}},\frac{\text{Lys Present}}{\text{Lys RDA}},\frac{\text{Met Present}}{\text{Met RDA}})\)
2.2 EAA-9 Calculation
We will be looking at four different calculations of the EAA-9 score:
- The Capped Average EAA-9 score is the average percent of the Recommended Dietary Allowance (RDA) of the 9 essential amino acids provided in 65 grams of protein where the maximum percent of the RDA that can be met is 100%
- The Uncapped Average EAA-9 score represents the average percent of the Recommended Dietary Allowance (RDA) of the 9 essential amino acids provided in 65 grams of protein
- The Capped minimum EAA-9 score is the minimum percentage of the RDA met by the 9 essential amino acids; capped at 100%
- The Uncapped minimum EAA-9 score is the minimum percentage of the RDA met by the 9 essential amino acids
Capped Average EAA-9:
\(\text{EAA-9}=\frac{(Min(\frac{\text{His Present}}{\text{His RDA}},1)+Min(\frac{\text{Ile Present}}{\text{Ile RDA}},1)+Min(\frac{\text{Leu Present}}{\text{Leu RDA}},1)+Min(\frac{\text{Lys Present}}{\text{Lys RDA}},1)+Min(\frac{\text{Met Present}}{\text{Met RDA}},1)+Min(\frac{\text{Phe Present}}{\text{Phe RDA}},1)+Min(\frac{\text{Thr Present}}{\text{Thr RDA}},1)+Min(\frac{\text{Trp Present}}{\text{Trp RDA}},1)+Min(\frac{\text{Val Present}}{\text{Val RDA}},1))}{9}\)
Uncapped Average EAA-9:
\(\text{EAA-9} = \frac{(\frac{\text{His Present}}{\text{His RDA}}+\frac{\text{Ile Present}}{\text{Ile RDA}}+\frac{\text{Leu Present}}{\text{Leu RDA}}+\frac{\text{Lys Present}}{\text{Lys RDA}}+\frac{\text{Met Present}}{\text{Met RDA}}+\frac{\text{Phe Present}}{\text{Phe RDA}}+\frac{\text{Thr Present}}{\text{Thr RDA}}+\frac{\text{Trp Present}}{\text{Trp RDA}}+\frac{\text{Val Present}}{\text{Val RDA}})}{9}\)
Capped Minimum EAA-9:
\(\text{EAA-9}=Min(Min(\frac{\text{His Present}}{\text{His RDA}},1),Min(\frac{\text{Ile Present}}{\text{Ile RDA}},1),Min(\frac{\text{Leu Present}}{\text{Leu RDA}},1),Min(\frac{\text{Lys Present}}{\text{Lys RDA}},1),Min(\frac{\text{Met Present}}{\text{Met RDA}},1),Min(\frac{\text{Phe Present}}{\text{Phe RDA}},1),Min(\frac{\text{Thr Present}}{\text{Thr RDA}},1),Min(\frac{\text{Trp Present}}{\text{Trp RDA}},1),Min(\frac{\text{Val Present}}{\text{Val RDA}},1))\)
Uncapped Minimum EAA-9:
\(\text{EAA-9}=Min(\frac{\text{His Present}}{\text{His RDA}},\frac{\text{Ile Present}}{\text{Ile RDA}},\frac{\text{Leu Present}}{\text{Leu RDA}},\frac{\text{Lys Present}}{\text{Lys RDA}},\frac{\text{Met Present}}{\text{Met RDA}},\frac{\text{Phe Present}}{\text{Phe RDA}},\frac{\text{Thr Present}}{\text{Thr RDA}},\frac{\text{Trp Present}}{\text{Trp RDA}},\frac{\text{Val Present}}{\text{Val RDA}})\)
2.3 FAO Amino Acid Score Calculation
The FAO Amino Acid Score is a large component of calculating PDCAAS. This score (as provided in the Report of the Joint FAO/WHO Expert Consultation on Protein Quality Evaluation (2007)) is calculated as the proportion of the limiting amino acid in 1 gram of protein to a reference pattern. Where the reference pattern is indicated by the amount of that amino acid present in 1 gram of protein from breast milk.
\(\scriptstyle \text{Amino Acid Score}=Min(Min(\frac{\text{His Present*}}{\text{His Scoring Pattern}},1),Min(\frac{\text{Ile Present*}}{\text{Ile Scoring Pattern}},1),Min(\frac{\text{Leu Present*}}{\text{Leu Scoring Pattern}},1),Min(\frac{\text{Lys Present*}}{\text{Lys Scoring Pattern}},1),Min(\frac{\text{Met Present*}}{\text{Met Scoring Pattern}},1),Min(\frac{\text{Phe Present*}}{\text{Phe Scoring Pattern}},1),Min(\frac{\text{Thr Present*}}{\text{Thr Scoring Pattern}},1),Min(\frac{\text{Trp Present*}}{\text{Trp Scoring Pattern}},1),Min(\frac{\text{Val Present*}}{\text{Val Scoring Pattern}},1))\)
\(\scriptstyle \text{*Per 1 gram of protein}\)
The following figures are taken directly from the Report of the Joint FAO/WHO Expert Consultation on Protein Quality Evaluation (2007)).
Amino Acid Scoring Patterns (FAO 2007):
Amino Acid Score as defined by FAO:
3 Validation: Amino Acid Score Comparisons
Why PDCAAS?
The international standard for protein quality is called PDCAAS (ie. Protein Digestibility Corrected Amino Acid Score) and established by the Food and Agriculture Organization (FAO). For validation, we will prove that our frameworks are consistent with this international standard.
PDCAAS is a function of an amino acid score (based on amino acid concentrations in breast milk) and protein digestibility. This dependence on protein digestibility severely limits the scope of usage of PDCAAS. The USDA is the gold standard of food and nutrient data, yet we are only able to calculate PDCAAS for 1% (85/7793) of SR Legacy entries. The vast food system includes upwards of a million foods, if we look at just grocery items provided by Nutritionix PDCAAS scores can only be calculated for .01% of the foods consumed every day.
Since protein digestibility and amino acid score are independent, validating an amino acid-based protein score against PDCAAS requires validating the score against the amino acid score used to calculate PDCAAS. Once the relationship between the amino acid scores is confirmed, we can calculate two versions of the PDCAAS (one using each amino acid score) and directly compare the two PDCAAS scores.
Forming A Direct Comparison
The FAO amino acid score determines the effectiveness with which absorbed dietary nitrogen can meet indispensable amino acid requirement at the safe level of protein intake using a comparison of the content of the limiting amino acid in the protein with its content in the requirement pattern. The EAA score is a measure of effectiveness of meeting the indispensable amino acid requirement for an individual based on their weight and the portion of food consumed. To form a valid comparison of the two amino acid scoring systems, we need to define the constraints necessary for both scoring systems to measure equivalent concepts.
FAO amino acid score measures effectiveness per gram of protein by assuming a minimum requirement of protein intake, meaning we have to scale the EAA score to a minimum requirement of protein intake to form a direct comparison. EAA is defined in terms of the Institute of Medicine DRI reports, in which a minimum requirement of protein is initially defined for an adult male weighing 70kg as 56 grams per day of protein and then subsequently scaled by kg to 0.8 grams of protein per kg per day. Therefore, calculating EAA score for a 70kg man of foods containing 56 grams of protein is equivalent to measuring the effectiveness per gram of protein by assuming a minimum requirement of protein intake. This allows for a one-to-one comparison of EAA and FAO amino acid scores.
Foods used for Comparisons
To assess protein quality we need to compare foods containing enough protein to reasonably meet amino acid requirements. According to MyPlate meats, poultry, seafood, eggs, nuts and seeds, and beans, peas, and lentils are all “Protein Food Groups”.
To validate our scoring frameworks against the current standards, we will use the 2956 SR Legacy foods (for which information on all essential amino acids are present) in the SR categories: “Dairy and Egg Products”, “Pork Products”, “Finfish and Shellfish Products”, “Legumes and Legume Products”, “Nut and Seed Products”, “Beef Products”, “Sausages and Luncheon Meats”, “Poultry Products”, and “Lamb, Veal, and Game Products”.
Statistical Methodology
To ascertain the degree to which the scores evaluate a food in a parallel manner, an assessment of rank and variability is required rather than an assessment of value (i.e. beef should be rated higher than corn regardless of score values). Two statistical tests are implemented here toward that goal:
- Spearman Ranked Correlation
- Spearman’s Rank correlation coefficient is a technique which can be used to summarise the strength and direction (negative or positive) of a relationship between two variables.
- The correlation coefficient measures the degree to which the two scores rank the foods in the same order where a 1 (or 100%) would indicated that two scores rank all foods identically and 0 would indicate no correlation between the ranks of the two scores.
- Fligner-Killeen Test
- This test determines the homogeneity of variances (i.e. whether or not one score is more variable than another)
- Variability indicates the spread of a score. If two scores have high rank correlation, a score with higher variability will have values that are further spread apart and easier to distinguish.
In addition, there is a paired plot for each score comparison to visually identify any trends in discrepancies between scores and visually confirm conclusions from statistical tests.
3.1 Capped Average
Capped Average EAA-3 vs FAO
Paired Plot
Figure 1: Comparison of Amino Acid Scoring Methods
Key notes/important takeaways:
- FAO Amino Acid Score is occasionally extremely lower than Capped EAA-3 Score. This happens when a food has little to no Histidine, Isoleucine, Phenylalanine, Threonine, Tryptophan, and/or Valine but relatively large amounts of Leucine, Lysine, and Methionine.
Spearman Rank Correlation
##
## Spearman's rank correlation rho
##
## data: AA_Score_Comparison$Protein_Quality_Index and AA_Score_Comparison$EAA3
## S = 1138941690, p-value < 2.2e-16
## alternative hypothesis: true rho is not equal to 0
## sample estimates:
## rho
## 0.7354307Key notes/important takeaways:
- Since p-value is < 2.2e-16, we can conclude that the two scores are correlated
- The correlation coefficient is 73.5%
Fligner Test (homogeneity of variance)
##
## Fligner-Killeen test of homogeneity of variances
##
## data: score by Method
## Fligner-Killeen:med chi-squared = 71.581, df = 1, p-value < 2.2e-16SD of FAO Amino Acid Score:
## [1] 0.1431547SD of Capped Average EAA-3:
## [1] 0.05960405Key notes/important takeaways:
- Since p-value is < 2.2e-16, we can conclude that the variance of the two scores is significantly different.
- Capped Average EAA-3 is less variable than FAO Amino Acid Score (SD of FAO Amino Acid Score is 8.4% higher).
Capped Average EAA-9 vs FAO
Paired Plot
Figure 2: Comparison of Amino Acid Scoring Methods
Key notes/important takeaways:
- Capped Average EAA-3 is a better approximation of FAO amino acid score than Capped Minimum EAA-9.
Spearman Rank Correlation
##
## Spearman's rank correlation rho
##
## data: AA_Score_Comparison$Protein_Quality_Index and AA_Score_Comparison$EAA9
## S = 1.009e+09, p-value < 2.2e-16
## alternative hypothesis: true rho is not equal to 0
## sample estimates:
## rho
## 0.7656087Key notes/important takeaways:
- Since p-value is < 2.2e-16, we can conclude that the two scores are correlated
- The correlation coefficient is 76.5%
Fligner Test (homogeneity of variance)
##
## Fligner-Killeen test of homogeneity of variances
##
## data: score by Method
## Fligner-Killeen:med chi-squared = 334.57, df = 1, p-value < 2.2e-16SD of FAO Amino Acid Score:
## [1] 0.1431547SD of EAA-9:
## [1] 0.02897123Key notes/important takeaways:
- Since p-value is < 2.2e-16, we can conclude that the variance of the two scores is significantly different.
- FAO Amino Acid Score is more variable than Capped Average EAA-9 score (SD of FAO Amino Acid Score is 11.4% higher).
Capped Average EAA-3 vs Capped Average EAA-9
Paired Plot
Figure 3: Comparison of Amino Acid Scoring Methods
Key notes/important takeaways:
- Capped Average EAA-9 is often higher than Capped Average EAA-3 since Leucine, Lysine, and Methionine are often limiting amino acids.
- Capped Average EAA-9 is lower than Capped Average EAA-3 when Histidine, Isoleucine, Phenylalanine, Threonine, Tryptophan, or Valine are limiting amino acids. Phenylalanine is limiting most often in this case.
Spearman Rank Correlation
##
## Spearman's rank correlation rho
##
## data: AA_Score_Comparison$EAA9 and AA_Score_Comparison$EAA3
## S = 1708869038, p-value < 2.2e-16
## alternative hypothesis: true rho is not equal to 0
## sample estimates:
## rho
## 0.60304Key notes/important takeaways:
- Since p-value is < 2.2e-16, we can conclude that the two scores are correlated
- The correlation coefficient is 60.3%
Fligner Test (homogeneity of variance)
##
## Fligner-Killeen test of homogeneity of variances
##
## data: score by Method
## Fligner-Killeen:med chi-squared = 671.74, df = 1, p-value < 2.2e-16SD of Capped Average EAA-9:
## [1] 0.02897123SD of Capped Average EAA-3:
## [1] 0.05960405Key notes/important takeaways:
- Since p-value is 1.03e-05, we can conclude that the variance of the two scores is significantly different.
- Capped Average EAA-3 is more variable than Capped Average EAA-9 (SD of Capped Average EAA-9 is 3.1% higher).
3.2 Uncapped Average
Uncapped Average EAA-3 vs FAO
Paired Plot
Figure 4: Comparison of Amino Acid Scoring Methods
Key notes/important takeaways:
- Large differences between Uncapped Average EAA-3 scores FAO amino acid scores are present for foods in which the limiting amino acid is not leucine, lysine or methionine.
- The capped average is a better approximation than uncapped average
Spearman Rank Correlation
##
## Spearman's rank correlation rho
##
## data: AA_Score_Comparison$Protein_Quality_Index and AA_Score_Comparison$EAA3
## S = 1452345363, p-value < 2.2e-16
## alternative hypothesis: true rho is not equal to 0
## sample estimates:
## rho
## 0.6626289Key notes/important takeaways:
- Since p-value is < 2.2e-16, we can conclude that the two scores are correlated
- The correlation coefficient is 66.3%
Fligner Test (homogeneity of variance)
##
## Fligner-Killeen test of homogeneity of variances
##
## data: score by Method
## Fligner-Killeen:med chi-squared = 1189.4, df = 1, p-value < 2.2e-16SD of FAO Amino Acid Score:
## [1] 0.1431547SD of Uncapped Average EAA-3:
## [1] 0.2058319Key notes/important takeaways:
- Since p-value is < 2.2e-16, we can conclude that the variance of the two scores is significantly different.
- Uncapped Average EAA-3 is less variable than FAO Amino Acid Score (SD of Uncapped Average EAA-3 is higher by 6.3%).
Key notes/important takeaways:
- We can reject null hypothesis and conclude that the FAO Amino Acid Score is significantly different from the Uncapped Average EAA-3 score with a p-value < 2.2e-16.
- We are 95% confident that the Uncapped Average EAA-3 score is between 50.7% and 51.7% lower than the FAO Amino Acid Score.
Uncapped Average EAA-9 vs FAO
Paired Plot
Figure 5: Comparison of Amino Acid Scoring Methods
Key notes/important takeaways:
- Large differences between these scores are primarily due to differences between RDAs and FAO amino acid scoring patterns
Spearman Rank Correlation
##
## Spearman's rank correlation rho
##
## data: AA_Score_Comparison$Protein_Quality_Index and AA_Score_Comparison$EAA9
## S = 1811732172, p-value < 2.2e-16
## alternative hypothesis: true rho is not equal to 0
## sample estimates:
## rho
## 0.5791455Key notes/important takeaways:
- Since p-value is < 2.2e-16, we can conclude that the two scores are correlated
- The correlation coefficient is 57.9%
Fligner Test (homogeneity of variance)
##
## Fligner-Killeen test of homogeneity of variances
##
## data: score by Method
## Fligner-Killeen:med chi-squared = 1042.1, df = 1, p-value < 2.2e-16SD of FAO Amino Acid Score:
## [1] 0.1431547SD of Uncapped Average EAA-9:
## [1] 0.1816088Key notes/important takeaways:
- Since p-value is < 2.2e-16, we can conclude that the variance of the two scores is significantly different.
- Uncapped Average EAA-9 is more variable than FAO Amino Acid Score (SD of Uncapped Average EAA-9 is higher by 3.8%).
Uncapped Average EAA-3 vs Uncapped Average EAA-9
Paired Plot
Figure 6: Comparison of Amino Acid Scoring Methods
Key notes/important takeaways:
Spearman Rank Correlation
##
## Spearman's rank correlation rho
##
## data: AA_Score_Comparison$EAA9 and AA_Score_Comparison$EAA3
## S = 1220158584, p-value < 2.2e-16
## alternative hypothesis: true rho is not equal to 0
## sample estimates:
## rho
## 0.7165645Key notes/important takeaways:
- Since p-value is < 2.2e-16, we can conclude that the two scores are correlated
- The correlation coefficient is 71.7%
Fligner Test (homogeneity of variance)
##
## Fligner-Killeen test of homogeneity of variances
##
## data: score by Method
## Fligner-Killeen:med chi-squared = 18.136, df = 1, p-value = 2.056e-05SD of Uncapped Average EAA-9:
## [1] 0.1816088SD of Uncapped Average EAA-3:
## [1] 0.2058319Key notes/important takeaways:
- Since p-value is 2.056e-05, we can conclude that the variance of the two scores is significantly different.
- Uncapped Average EAA-3 is more variable than Uncapped Average EAA-9 (SD of Uncapped Average EAA-3 is higher by 2.4%).
3.3 Capped Minimum
Capped Minimum EAA-3 vs FAO
Paired Plot
Figure 7: Comparison of Amino Acid Scoring Methods
Key notes/important takeaways:
- FAO Amino Acid Score is much lower than Capped Minimum EAA-3 score when the limiting amino acid in a food is not leucine, lysine, or methionine.
Spearman Rank Correlation
##
## Spearman's rank correlation rho
##
## data: AA_Score_Comparison$Protein_Quality_Index and AA_Score_Comparison$EAA3
## S = 1137564244, p-value < 2.2e-16
## alternative hypothesis: true rho is not equal to 0
## sample estimates:
## rho
## 0.7357506Key notes/important takeaways:
- Since p-value is < 2.2e-16, we can conclude that the two scores are correlated
- The correlation coefficient is 73.6%
Fligner Test (homogeneity of variance)
##
## Fligner-Killeen test of homogeneity of variances
##
## data: score by Method
## Fligner-Killeen:med chi-squared = 3.5838, df = 1, p-value = 0.05835SD of FAO Amino Acid Score:
## [1] 0.1431547SD of Capped Minimum EAA-3:
## [1] 0.1446254Key notes/important takeaways:
- Since p-value is 0.05835, we can conclude that the variance of the two scores is NOT significantly different.
Capped Minimum EAA-9 vs FAO
Paired Plot
Figure 8: Comparison of Amino Acid Scoring Methods
Key notes/important takeaways:
- Differences between these scores are present due to differences between RDA and amino acid scoring patterns.
Spearman Rank Correlation
##
## Spearman's rank correlation rho
##
## data: AA_Score_Comparison$Protein_Quality_Index and AA_Score_Comparison$EAA9
## S = 1001829244, p-value < 2.2e-16
## alternative hypothesis: true rho is not equal to 0
## sample estimates:
## rho
## 0.7672811Key notes/important takeaways:
- Since p-value is < 2.2e-16, we can conclude that the two scores are correlated
- The correlation coefficient is 76.7%
Fligner Test (homogeneity of variance)
##
## Fligner-Killeen test of homogeneity of variances
##
## data: score by Method
## Fligner-Killeen:med chi-squared = 821.26, df = 1, p-value < 2.2e-16SD of FAO Amino Acid Score:
## [1] 0.1431547SD of Capped Minimum EAA-9:
## [1] 0.1476634Key notes/important takeaways:
- Since p-value is < 2.2e-16, we can conclude that the variance of the two scores is significantly different.
- FAO Amino Acid Score is less variable than Capped Minimum EAA-9 score (SD of Capped Minimum EAA-9 is 0.5% higher).
Capped Minimum EAA-3 vs Capped Minimum EAA-9
Paired Plot
Figure 9: Comparison of Amino Acid Scoring Methods
Key notes/important takeaways:
- Capped Minimum EAA-9 Amino Acid Score is much lower than Capped Minimum EAA-3 score when the limiting amino acid in a food is not leucine, lysine, or methionine.
Spearman Rank Correlation
##
## Spearman's rank correlation rho
##
## data: AA_Score_Comparison$EAA9 and AA_Score_Comparison$EAA3
## S = 1829535230, p-value < 2.2e-16
## alternative hypothesis: true rho is not equal to 0
## sample estimates:
## rho
## 0.5750099Key notes/important takeaways:
- Since p-value is < 2.2e-16, we can conclude that the two scores are correlated
- The correlation coefficient is 57.5%
Fligner Test (homogeneity of variance)
##
## Fligner-Killeen test of homogeneity of variances
##
## data: score by Method
## Fligner-Killeen:med chi-squared = 1056.3, df = 1, p-value < 2.2e-16SD of Capped Minimum EAA-9:
## [1] 0.1476634SD of Capped Minimum EAA-3:
## [1] 0.1446254Key notes/important takeaways:
- Since p-value is < 2.2e-16, we can conclude that the variance of the two scores is significantly different.
- Capped Minimum EAA-3 is less variable than Capped Minimum EAA-9 (SD of Capped Minimum EAA-9 is 0.3% higher).
3.4 Uncapped Minimum
Uncapped Minimum EAA-3 vs FAO
Paired Plot
Figure 10: Comparison of Amino Acid Scoring Methods
Key notes/important takeaways:
- FAO Amino Acid Score is much lower than Uncapped Minimum EAA-3 score when the limiting amino acid in a food is not leucine, lysine, or methionine.
Spearman Rank Correlation
##
## Spearman's rank correlation rho
##
## data: AA_Score_Comparison$Protein_Quality_Index and AA_Score_Comparison$EAA3
## S = 1652159675, p-value < 2.2e-16
## alternative hypothesis: true rho is not equal to 0
## sample estimates:
## rho
## 0.6162132Key notes/important takeaways:
- Since p-value is < 2.2e-16, we can conclude that the two scores are correlated
- The correlation coefficient is 61.6%
Fligner Test (homogeneity of variance)
##
## Fligner-Killeen test of homogeneity of variances
##
## data: score by Method
## Fligner-Killeen:med chi-squared = 1278.1, df = 1, p-value < 2.2e-16SD of FAO Amino Acid Score:
## [1] 0.1431547SD of Uncapped Minimum EAA-3:
## [1] 0.2247256Key notes/important takeaways:
- Since p-value is < 2.2e-16, we can conclude that the variance of the two scores is significantly different.
- Uncapped Minimum EAA-3 is more variable than FAO Amino Acid Score (SD of Uncapped Minimum EAA-3 is 8.23% higher).
Uncapped Minimum EAA-9 vs FAO
Paired Plot
Figure 11: Comparison of Amino Acid Scoring Methods
Key notes/important takeaways:
- Differences between these scores are present due to differences between RDA and amino acid scoring patterns.
Spearman Rank Correlation
##
## Spearman's rank correlation rho
##
## data: AA_Score_Comparison$Protein_Quality_Index and AA_Score_Comparison$EAA9
## S = 1.018e+09, p-value < 2.2e-16
## alternative hypothesis: true rho is not equal to 0
## sample estimates:
## rho
## 0.7635225Key notes/important takeaways:
- Since p-value is < 2.2e-16, we can conclude that the two scores are correlated
- The correlation coefficient is 76.35%
Fligner Test (homogeneity of variance)
##
## Fligner-Killeen test of homogeneity of variances
##
## data: score by Method
## Fligner-Killeen:med chi-squared = 942.83, df = 1, p-value < 2.2e-16SD of FAO Amino Acid Score:
## [1] 0.1431547SD of Uncapped Minimum EAA-9:
## [1] 0.1703788Key notes/important takeaways:
- Since p-value is < 2.2e-16, we can conclude that the variance of the two scores is significantly different.
- FAO Amino Acid Score is less variable than Uncapped Minimum EAA-9 score (SD of Uncapped Minimum EAA-9 is 2.7% higher).
Uncapped Minimum EAA-3 vs Uncapped Minimum EAA-9
Paired Plot
Figure 12: Comparison of Amino Acid Scoring Methods
Key notes/important takeaways:
- Uncapped Minimum EAA-9 Amino Acid Score is much lower than Uncapped Minimum EAA-3 score when the limiting amino acid in a food is not leucine, lysine, or methionine.
Spearman Rank Correlation
##
## Spearman's rank correlation rho
##
## data: AA_Score_Comparison$EAA9 and AA_Score_Comparison$EAA3
## S = 2396414548, p-value < 2.2e-16
## alternative hypothesis: true rho is not equal to 0
## sample estimates:
## rho
## 0.4433273Key notes/important takeaways:
- Since p-value is < 2.2e-16, we can conclude that the two scores are correlated
- The correlation coefficient is 44.33%
Fligner Test (homogeneity of variance)
##
## Fligner-Killeen test of homogeneity of variances
##
## data: score by Method
## Fligner-Killeen:med chi-squared = 246.65, df = 1, p-value < 2.2e-16SD of Uncapped Minimum EAA-9:
## [1] 0.1703788SD of Uncapped Minimum EAA-3:
## [1] 0.2247256Key notes/important takeaways:
- Since p-value is < 2.2e-16, we can conclude that the variance of the two scores is significantly different.
- Uncapped Minimum EAA-3 is more variable than Uncapped Minimum EAA-9 (SD of Uncapped Minimum EAA-3 is 5.4% higher).
4 Validation: PDCAAS Calculations and Comparison
PDCAAS is calculated using the following formula:
\[PDCAAS = \text{Amino Acid Score} * \text{Protein Digestibility}\]
Where the amino acid score is calculated as the proportion of the limiting amino acid in 1 gram of protein to a reference pattern. Where the reference pattern is indicated by the amount of that amino acid present in 1 gram of protein from breast milk (as seen above in section 3.3).
PDCAAS calculation requires the use of protein digestibility scores, which can often be difficult to aggregate. We collected protein digestibility scores from the Genesis R&D Food Manual which provides ~220 protein digestibility scores. We were then able to identify 96 foods in SR Legacy with full amino acid profiles that could be mapped to the available protein digestibility scores.
After identifying these 96 foods, we calculated 2 versions of the PDCAAS; one using the amino acid score proposed by the FAO and one using the Uncapped Minimum EAA-3 score.
Since the Uncapped Minimum EAA-3 score closely approximates the FAO Amino Acid Score, PDCAAS scores calculated using the Uncapped Minimum EAA-3 score should closely approximate PDCAAS scores calculated using the FAO Amino Acid Score.
Paired Plot
Figure 13: Comparison of PDCAAS
Spearman Rank Correlation
##
## Spearman's rank correlation rho
##
## data: PDCAAS_scores$FAO_PDCAAS and PDCAAS_scores$EAA3_PDCAAS
## S = 2134.3, p-value < 2.2e-16
## alternative hypothesis: true rho is not equal to 0
## sample estimates:
## rho
## 0.955447Key notes/important takeaways:
- Since p-value is < 2.2e-16, we can conclude that the two scores are correlated
- The correlation coefficient is 95.5%!!!!
Fligner Test (homogeneity of variance)
##
## Fligner-Killeen test of homogeneity of variances
##
## data: score by Method
## Fligner-Killeen:med chi-squared = 14.564, df = 1, p-value = 0.0001355SD of PDCAAS scores calculated using FAO Amino Acid Score:
## [1] 0.2395406SD of PDCAAS scores calculated using Capped Minimum EAA-3:
## [1] 0.3595222Key notes/important takeaways:
- Since p-value is 0.0001355, we can conclude that the variance of the two scores is significantly different
- EAA-3 PDCAAS is more variable by 12%
Figure 14: Overall Comparison of PDCAAS
Option 1
Option 2
Option 3
Key notes/important takeaways:
- PDCAAS is calculated as a product of amino acid score and protein digestibility. If you calculate the PDCAAS using an un-capped (not truncated) amino acid score then the PDCAAS will also be un-capped. This is why the animal-based foods that meet amino acid requirements in less than 56 grams of protein have much larger PDCAAS values when calculated using the EAA-3 score.
- For example, the food with highest EAA-3 PDCAAS is raw egg which has an EAA-3 score above 150%
- The plant-based foods have lower PDCAAS values when calculated using the EAA-3 score rather than the FAO score because the EAA-3 score is calculated using the IOM requirements as opposed to the FAO requirements. IOM has higher recommendations for leucine and lysine, which are the most frequent limiting amino acids in plant-based foods.
- For many foods (particularly plant-based foods), there is little difference between the different PDCAAS values. For instance, cheddar cheese and chickpeas both are consistent across scoring methods.
5 Validation Summary
The Uncapped and Capped Minimum Score Frameworks closely approximate the FAO amino acid score. While the correlation is significantly higher for the Uncapped Min EAA-9 and FAO than the Uncapped Min EAA-3 and FAO, this does not necessarily imply that EAA-9 is superior to EAA-3.
In addition, FAO has decided to move toward the use of uncapped or non-truncated scores going forward. Therefore there is no need to justify an established decision.
Since the EAA-3 frameworks focus on the 3 amino acids most essential to maintaining health, they are more valuable measures of protein quality. The EAA-9 frameworks are always going to be more representative of the FAO score since both rely on all 9 essential amino acids.
Using a score that is more closely associated with human health, that requires less data to calculate than other available measures could have a tremendous positive impact.
We are 95% confident there is a strong positive monotonic relationship between the Minimum EAA scores (all p-values < 2.2e-16) and FAO amino acid score as well as a strong positive monotonic relationship between PDCAAS values calculated using the two amino acid scoring systems.
With the EAA frameworks sufficiently validated against international standards, we can move toward applying them in a meaningful context. We propose a series of interventions to be undertaken as soon as feasible to help ensure that protein quality can be accurately assessed by the general public and avoid the negative health repercussions of amino acid deficiency.
6 Consumer Applications
6.1 Building Amino Acid Complete Diets
This scoring framework is particularly valuable since it is scalable to meals. For instance, suppose that you wanted to find the Uncapped Minimum EAA-3 score of a turkey and cheese sandwich. The calculation would look like this:
\(\text{EAA-3}= Min(\frac{\text{Leu in turkey + Leu in cheese + Leu in bread}}{\text{Leu RDA}},\frac{\text{Lys in turkey + Lys in cheese + Lys in bread}}{\text{Lys RDA}},\frac{\text{Met in turkey + Met in cheese + Met in bread}}{\text{Met RDA}})\)
The Uncapped Minimum EAA-3 score can then be used to compare meals and portion sizes.
Figure 15: EAA-3 Score Meal Comparisons by Serving Size
Figure 16: EAA-3 Score Meal Comparisons by MyPlate Oz Equivalents
Since the Uncapped Minimum EAA-3 score is additive, if the sum of scores for the meals you consume in one day reaches 100% you are guaranteed to have met amino acid requirements for that day for leucine, lysine, and methionine and 95% confident to have met requirements for all 9 amino acids. However, if any two meals consumed in one day have different limiting amino acids, you may meet requirements before the sum of scores reaches 100%.
By this logic, you’ll notice from the above graphic that it is very easy to meet amino acid requirements using animal-based proteins and very difficult to meet requirements using plant-based proteins. So what combinations of plant-based proteins would result in meeting your daily requirements?
Plant-based foods are most commonly limited by their amount of lysine and methionine. By complimenting foods containing limiting amounts of lysine such as wild rice or sesame seeds with foods containing limiting amounts of methionine such as soybeans or tofu, it is possible to meet amino acid requirements. This assessment of complementary pairings can be easily tracked using the EAA-3 score (i.e. the combined score of two complementary foods will be higher than adding the individual scores of each food).
6.2 Oz Equivalents
Ounce equivalents are defined on the MyPlate website using the “Ounce-Equivalent of Protein Foods Table.”
“This chart lists specific amounts that count as 1 ounce-equivalent (oz-equiv) in the Protein Foods Group towards your daily recommended intake:”
Evaluation of Current Oz Equivalents
To evaluate the extent to which the protein provided by the oz equivalents is truly equivalent, we will compare the amino acid contents of the defined portions.
FDC IDS of Individual foods used for comparisons:
- 170567
- 171140
- 173735
- 170557
Figure 17: Amino Acid Content of Ounce Equivalents
Figure 18: Ounce Equivalents Needed to Meet RDA
Figure 19: Average Amino Acid Content per Ounce Equivalent by Protein Food Group
Figure 20: Protein Group Ounce Equivalents Needed to Meet RDA
FDC IDs of foods used in averages can be found in table 5
Creation of New Oz Equivalents
The following is a proposed methodology for calculating accurate oz equivalents between food groups.
Egg is a standard representation of a complete protein, thus we will calculate the amount of each food needed to fulfill the same protein requirements as 1 egg. The amino acid content of an egg is taken from a 50g hard-boiled egg in SR Legacy (FDC ID: 173424 ). The limiting amino acid in 1 large (50g) hard boiled egg is phenylalanine at 668mg; for a 70 kg man this would fulfill 28.92% of the RDA. To calculate the “real-equivalents” to 1 egg, we will use the amount in grams of each food needed to acquire 28.92% or more of RDA for its limiting amino acid.
To determine the amino acid content of each food in the MyPlate oz equivalents, foods in SR Legacy with full amino acid profiles were matched to the food descriptions in the MyPlate Protein Foods table. If more than one food in SR Legacy matched the description, the average amino acid content of the foods was used.
7 Research Applications
7.1 Standard Amino Acid Recommendations
Recommended intake of amino acid varies across sources. The Uncapped EAA-3 score framework can be used to calculate protein quality using any of these requirements. For example, we will compare the Uncapped Minimum EAA-3 calculated using the FAO estimated amino acid requirements, the IOM RDAs, the IAAO recommendations, and the 95% CI upper limit of the IAAO recommendations.
While we have chosen to showcase the scores using these recommendations, any amino acid score recommendations can be used for calculation.
For this example, we will compare scores for a 70kg man. That being said, the scoring framework can be scaled for any weight, age range, or sex. If a recommendation of daily intake of amino acid can be identified for a specific group, the framework can be used to create specific scores associated with those requirements.
7.2 Optimizing outcomes: Beyond Min Requirements
Leucine is essential to regulation of muscle protein synthesis, insulin signaling and glucose re-cycling via alanine. It is estimated that the stimulation of muscle protein synthesis would be optimized with 18 g IAA, including 2.5 g leucine, at each of 3 meals per day (source).
In the following, we calculated EAA-3 scores using 7.5 g of leucine for muscle protein synthesis. Values for methionine and lysine are given by the IAAO 95% CI upper bound provided above.
While this example uses a recommendation for leucine independent of age, if a different leucine requirement can be identified for each age group then a different EAA-3 score can be provided for each age group.
7.3 Assessment of Protein per 100kcal
Scores such as the Nutrient-Rich Food (NRF) score are being used to calculate overall nutrient density of foods based on caloric intake. The EAA score can be used in a similar manner to evaluate the protein density of foods. By calculating the EAA-3 score of 100kcal portions of food, we can determine their relative protein quality by calorie.
The following table portrays the EAA-3 score calculated for a 70kg man using each of the recommendations provided above.
Figure 21: Overall Comparison of EAA-3 Scores By Requirement and Food Category
Figure 22: Comparison of EAA-3 Scores By Requirement and Food Category
8 Food Industry Applications
8.1 Front of Package Health Claims
The EAA-3 score as a tool to confirm and support front of label health claims.
- Structure / function
- “Good source of”
The EAA-3 framework can be used along protein characteristics to drive industry choices of protein. An example of comparable characteristics can be seen below (not actual data, for example only).
“If I were to strategize how to go about getting people interested, I’d start with the dairy,beef, pork, and poultry industries to tell them that THEIR protein is of a higher quality than plant-based proteins on their own, and start a protein war between the industries and then start educating the plant based protein industry as to how to blend plant proteins for a higher quality score.”
“On the EAA-3 score, it is relevant for industry if it changes something. Meaning, can I use a different protein blend and claim a higher amount? Or will it be cheaper? Or will it have better organoleptic properties? Or will it be easier to use? That sort of thing. If it doesn’t change something, preferably for the better, then it will be difficult to get traction from the industry in the near term.”
9 Impact of Applying the EAA-3 Framework
PDCAAS is the current protein quality standard but the usefulness of PDCAAS is severely limited by the amount of available protein digestibility scores.
The USDA is the gold standard of food and nutrient data, yet we are only able to calculate PDCAAS for 1% (85/7793) of Legacy entries. The vast food system includes upwards of a million foods, if we look at just grocery items provided by Nutritionix PDCAAS scores can only be calculated for .01% of the foods consumed every day.