Reduced Rank Regression Model Research Articles

Robust Estimation of Multivariate Time Series Data Based on Reduced Rank Model

ABSTRACTMultivariate time series analysis uncovers the intricate relationships among multiple variables, which plays a vital role in areas such as policy‐making and business decision‐making. This paper employs a reduced rank regression model to investigate a robust estimation method for multivariate time series data using an penalty. The goal is to achieve rapid parameter estimation while ensuring robustness in the analysis of time series data. This study provides a detailed description of the solution process and examines the theoretical properties of the proposed method. To evaluate its effectiveness, the proposed model is compared with full‐rank regression and the multivariate regression with covariance estimation (MRCE) method through simulations, as well as an analysis of the Sceaux household electric power consumption data. The results indicate that the proposed model performs well.

Experimental development of a method of short and medium-term photovoltaic generation forecasting using multivariate statistics and mathematical modeling

Uncertainties in photovoltaic solar energy production can make it challenging to dispatch energy into the electricity grid. Although photovoltaic generation storage solves this problem, a forecasting of the photovoltaic solar energy produced is necessary to control the energy injected into the grid. This article aims to develop the probabilistic methodology Reduced-Rank Regression (RRR) for forecasting photovoltaic generation in the short and medium terms. The RRR methodology forecasting uses the generation data of a grid-connected photovoltaic system. The proposed RRR model is simple, easy to access and apply, and does not use irradiance data. The model developed uses the multivariate statistical analysis technique. A advantage is that with a correlation with the performance indices of photovoltaic solar energy systems, the proposed method can be applied in any geographical location on the planet and with different photovoltaic solar energy systems. The application of the RRR methodology requires two searches/inputs. The first input is weather forecast data obtained from a weather forecasting platform, and the second is actual historical data on photovoltaic generation at the site where the method was developed. The proposed method was compared with the persistence method. Using a horizon of 1–10 h, the average monthly root mean square error for the RRR ranged from 7.3 % to 50.1 %. For the persistence method, the average monthly root mean square error ranged from 15.1 % to 65.0 %. Therefore, with the horizon of 24 h, the average monthly root mean square error for the RRR ranged from 4.5 % to 43.2 %. For the persistence method, the average monthly root mean square error ranged from 11.5 % to 75.0 %. We show experimentally that our method is competitive with the state-of-the-art in terms of obtaining photovoltaic generation forecasting without using solar radiation data.

Bayesian reduced rank regression models generalizable neural fingerprints that differentiate between individuals in magnetoencephalography data.

Recent magnetoencephalography (MEG) studies have reported that functional connectivity (FC) and power spectra can be used as neural fingerprints in differentiating individuals. Such studies have mainly used correlations between measurement sessions to distinguish individuals from each other. However, it has remained unclear whether such correlations might reflect a more generalizable principle of individually distinctive brain patterns. Here, we evaluated a machine-learning based approach, termed latent-noise Bayesian reduced rank regression (BRRR) as a means of modelling individual differences in the resting-state MEG data of the Human Connectome Project (HCP), using FC and power spectra as neural features. First, we verified that BRRR could model and reproduce the differences between metrics that correlation-based fingerprinting yields. We trained BRRR models to distinguish individuals based on data from one measurement and used the models to identify subsequent measurement sessions of those same individuals. The best performing BRRR models, using only 20 spatiospectral components, were able to identify subjects across measurement sessions with over 90% accuracy, approaching the highest correlation-based accuracies. Using cross-validation, we then determined whether that BRRR model could generalize to unseen subjects, successfully classifying the measurement sessions of novel individuals with over 80% accuracy. The results demonstrate that individual neurofunctional differences can be reliably extracted from MEG data with a low-dimensional predictive model and that the model is able to classify novel subjects.

Post-processing for Bayesian analysis of reduced rank regression models with orthonormality restrictions

Orthonormality constraints are common in reduced rank models. They imply that matrix-variate parameters are given as orthonormal column vectors. However, these orthonormality restrictions do not provide identification for all parameters. For this setup, we show how the remaining identification issue can be handled in a Bayesian analysis via post-processing the sampling output according to an appropriately specified loss function. This extends the possibilities for Bayesian inference in reduced rank regression models with a part of the parameter space restricted to the Stiefel manifold. Besides inference, we also discuss model selection in terms of posterior predictive assessment. We illustrate the proposed approach with a simulation study and an empirical application.

Trimethylamine N-oxide, β-alanine, tryptophan index, and vitamin B6-related dietary patterns in association with stroke risk

Background and aimsThe aim of this study was to examine the associations of dietary patterns derived by reduced-rank regression (RRR) model reflecting variation in novel biomarkers (trimethylamine N-oxide, β-alanine, tryptophan index, and vitamin B6) with stroke risk. Methods and resultsWe performed analyses based on a community-based cohort study “the Prospective Follow-up Study on Cardiovascular Morbidity and Mortality in China (PFS-CMMC)”. Factor loadings were calculated by RRR using 11 food groups collected via a validated food frequency questionnaire and the four response variables based on its nested case–control data (393 cases of stroke vs. 393 matched controls). Dietary pattern scores were derived by applying the factor loadings to the food groups in the entire cohort (n = 15,518). The associations of dietary pattern with the stroke risk were assessed using Cox proportional hazards models. The dietary pattern characterized with higher intakes of red meat and poultry but lower intakes of fresh vegetables, fresh fruits, and fish/seafoods were identified for further analyses. The hazard ratios (HR) for the highest vs. lowest quartile was 1.55 [95 % confidence interval (CI): 1.18–2.03, P trend = 0.001] for total stroke, 2.96 [95 % CI: 1.53–5.71, P trend <0.001] for non-ischemic stroke, after adjustment for sex, age, educational attainment, current smoking, current drinking, body mass index, total energy intake, family history of stroke, hypertension, diabetes, hyperlipidemia, and estimated glomerular filtration rate. ConclusionOur findings highlight the importance of limited meat intake and increased intakes of fresh vegetables, fruits, and fish/seafoods in the prevention of stroke among Chinese adults.

A new algorithm and a discussion about visualization for logistic reduced rank regression

Logistic reduced rank regression is a useful data analysis tool when we have multiple binary response variables and a set of predictors. In this paper, we describe logistic reduced rank regression and present a new majorization minimization algorithm for the estimation of model parameters. Furthermore, we discuss Type I and Type D triplots for visualizing the results of a logistic reduced rank regression model, compare them, and then develop a hybrid triplot using elements of both types. Two empirical data sets are analyzed. This analysis is used to (1) compare the new algorithm to an existing one in terms of speed; and (2) to show the hybrid triplot and its interpretation.

Dietary Inflammatory Patterns Are Associated With Serum TGs and Insulin in Adults: A Community-Based Study in Taiwan

BackgroundDietary patterns related to inflammation have become a focus of disease prevention but the patterns may vary among populations. ObjectivesThe study was conducted to determine Taiwanese dietary inflammatory patterns and evaluate their associations with biomarkers of lipid and glucose. MethodsData were taken from 5664 community-dwelling individuals aged ≥55 y recruited in 2009–2013 in the Healthy Aging Longitudinal Study in Taiwan (HALST). Dietary data were obtained from an FFQ. An empirical dietary inflammatory pattern (EDIP) was derived from reduced rank regression models that explained the serum high-sensitivity CRP, plasma IL-6, and TNF receptor 1. Cross-sectional associations between dietary scores and biomarkers of total cholesterol (TC); HDL cholesterol; LDL cholesterol; TG; and ratios of TG/HDL cholesterol, TG/TC, fasting glucose, insulin, and HbA1c were analyzed via multiple linear regression and adjusted for major confounders. The false-discovery rate (FDR)-adjusted P < 0.05 was considered statistically significant. Abdominal obesity was defined as a waist circumference of ≥90 cm for men and ≥80 cm for women. ResultsHigher EDIP-HALST scores were associated with higher TG (per score increment: 1.62%, 95% CI: 0.58%, 2.76%; PFDR = 0.01), TG/HDL cholesterol (2.01%, 95% CI: 0.67%, 3.37%; PFDR = 0.01), and TG/TC (1.42%, 95% CI: 0.41%, 2.43%; PFDR = 0.01) and nonlinearly associated with insulin, with those in the middle tertile had the highest serum insulin concentrations (means: 5.12 μIU/mL, 95% CI: 4.78, 5.78; PFDR = 0.04) in men, but not in women. No heterogeneity was detected between sexes. The associations with TG (1.23%, 95% CI: 0.19, 2.23%; Ptrend = 0.02), TG/HDL cholesterol (1.62%, 95% CI: 0.30%, 2.96%; Ptrend = 0.02), and TG/TC (1.11%, 95% CI: 0.11%, 2.13%; Ptrend = 0.03) were stronger in participants with abdominal obesity, but were borderline associated in participants with normal abdominal circumferences (all Ptrend = 0.05). ConclusionsInflammatory diets, as measured via EDIP-HALST, are associated with serum TG concentration, particularly in participants with abdominal obesity. These findings may suggest that developing disease prevention strategies using dietary inflammatory patterns may be different by populations.J Nutr 20xx;x:xx.

Genetic architecture of hippocampus subfields volumes in Alzheimer's disease.

The hippocampus is a heterogeneous structure, comprising histologically and functionally distinguishable hippocampal subfields. The volume reductions in hippocampal subfields have been demonstrated to be linked with Alzheimer's disease (AD). The aim of our study is to investigate the hippocampal subfields' genetic architecture based on the Alzheimer's Disease Neuroimaging Initiative (ADNI) data set. After preprocessing the downloaded genetic variants and imaging data from the ADNI database, a co-sparse reduced rank regression model was applied to analyze the genetic architecture of hippocampal subfields volumes. Homology modeling, docking, molecular dynamics simulations, and Co-IP experiments for protein-protein interactions were used to verify the function of target protein on hippocampal subfields successively. After that, the association analysis between the candidated genes on the hippocampal subfields volume and clinical scales were performed. The results of the association analysis revealed five unique genetic variants (e.g., ubiquitin-specific protease 10 [USP10]) changed in nine hippocampal subfields (e.g., the granule cell and molecular layer of the dentate gyrus [GC-ML-DG]). Among five genetic variants, USP10 had the strongest interaction effect with BACE1, which affected hippocampal subfields verified by MD and Co-IP experiments. The results of association analysis between the candidated genes on the hippocampal subfields volume and clinical scales showed that candidated genes influenced the volume and function of hippocampal subfields. Current evidence suggests that hippocampal subfields have partly distinct genetic architecture and may improve the sensitivity of the detection of AD.

The association between genetic variations and morphology-based brain networks changes in Alzheimer's disease.

Alzheimer's disease (AD) is a highly heritable disease. The morphological changes of cortical cortex (such as, cortical thickness and surface area) in AD always accompany by the change of the functional connectivity to other brain regions and influence the short- and long-range brain network connections, causing functional deficits of AD. In this study, the first hypothesis is that genetic variations might affect morphology-based brain networks, leading to functional deficits; the second hypothesis is that protein-protein interaction (PPI) between the candidate proteins and known interacting proteins to AD might exist and influence AD. 600 470 variants and structural magnetic resonance imaging scans from 175 AD patients and 214 healthy controls were obtained from the Alzheimer's Disease Neuroimaging Initiative-1 database. A co-sparse reduced-rank regression model was fit to study the relationship between non-synonymous mutations and morphology-based brain networks. After that, PPIs between selected genes and BACE1, an enzyme that was known to be related to AD, are explored by using molecular dynamics (MD) simulation and co-immunoprecipitation (Co-IP) experiments. Eight genes affecting morphology-based brain networks were identified. The results of MD simulation showed that the PPI between TGM4 and BACE1 was the strongest among them and its interaction was verified by Co-IP. Hence, gene variations influence morphology-based brain networks in AD, leading to functional deficits. This finding, validated by MD simulation and Co-IP, suggests that the effect is robust.

Differences in dietary patterns related to metabolic health by gut microbial enterotypes of Korean adults.

Diet has a profound impact on the progression of metabolic syndrome (MetS) into various diseases. The gut microbiota could modulate the effect of diet on metabolic health. We examined whether dietary patterns related to MetS differed according to gut microbial enterotypes among 348 Korean adults aged 18-60 years recruited between 2018∼2021 in a cross-sectional study. The enterotype of each participant was identified based on 16S rRNA gut microbiota data. The main dietary pattern predicting MetS (MetS-DP) of each enterotype was derived using reduced-rank regression (RRR) models. In the RRR models, 27 food group intakes assessed by a semi-quantitative food frequency questionnaire and MetS prediction markers including triglyceride to high-density lipoprotein cholesterol (TG/HDL) ratio and homeostatic model assessment for insulin resistance (HOMA-IR) were used as predictor and response variables, respectively. The MetS-DP extracted in Bacteroides enterotype (B-type) was characterized by high consumption of refined white rice and low consumption of eggs, vegetables, and mushrooms. The MetS-DP derived among Prevotella enterotype (P-type) was characterized by a high intake of sugary food and low intakes of bread, fermented legumes, and fermented vegetables. The MetS-DP of B-type was positively associated with metabolic unhealthy status (OR T3 vs. T1 = 3.5; 95% CI = 1.5-8.2), comparing the highest tertile to the lowest tertile. Although it was not significantly associated with overall metabolic unhealthy status, the MetS-DP of P-type was positively associated with hyperglycemia risk (OR T3 vs. T1 = 6.2; 95% CI = 1.6-24.3). These results suggest that MetS-DP may differ according to the gut microbial enterotype of each individual. If such associations are found to be causal, personalized nutrition guidelines based on the enterotypes could be recommended to prevent MetS.

A Poisson reduced-rank regression model for association mapping in sequencing data.

Single-cell RNA-sequencing (scRNA-seq) technologies allow for the study of gene expression in individual cells. Often, it is of interest to understand how transcriptional activity is associated with cell-specific covariates, such as cell type, genotype, or measures of cell health. Traditional approaches for this type of association mapping assume independence between the outcome variables (or genes), and perform a separate regression for each. However, these methods are computationally costly and ignore the substantial correlation structure of gene expression. Furthermore, count-based scRNA-seq data pose challenges for traditional models based on Gaussian assumptions. We aim to resolve these issues by developing a reduced-rank regression model that identifies low-dimensional linear associations between a large number of cell-specific covariates and high-dimensional gene expression readouts. Our probabilistic model uses a Poisson likelihood in order to account for the unique structure of scRNA-seq counts. We demonstrate the performance of our model using simulations, and we apply our model to a scRNA-seq dataset, a spatial gene expression dataset, and a bulk RNA-seq dataset to show its behavior in three distinct analyses. We show that our statistical modeling approach, which is based on reduced-rank regression, captures associations between gene expression and cell- and sample-specific covariates by leveraging low-dimensional representations of transcriptional states.

Stability Approach to Regularization Selection for Reduced-Rank Regression

The reduced-rank regression model is a popular model to deal with multivariate response and multiple predictors, and is widely used in biology, chemometrics, econometrics, engineering, and other fields. In the reduced-rank regression modeling, a central objective is to estimate the rank of the coefficient matrix that represents the number of effective latent factors in predicting the multivariate response. Although theoretical results such as rank estimation consistency have been established for various methods, in practice rank determination still relies on information criterion based methods such as AIC and BIC or subsampling based methods such as cross-validation. Unfortunately, the theoretical properties of these practical methods are largely unknown. In this article, we present a novel method called StARS-RRR that selects the tuning parameter and then estimates the rank of the coefficient matrix for reduced-rank regression based on the stability approach. We prove that StARS-RRR achieves rank estimation consistency, that is, the rank estimated with the tuning parameter selected by StARS-RRR is consistent to the true rank. Through a simulation study, we show that StARS-RRR outperforms other tuning parameter selection methods including AIC, BIC, and cross-validation as it provides the most accurate estimated rank. In addition, when applied to a breast cancer dataset, StARS-RRR discovers a reasonable number of genetic pathways that affect the DNA copy number variations and results in a smaller prediction error than the other methods with a random-splitting process. Supplementary materials for this article are available online.

LARGE-SCALE MULTIVARIATE SPARSE REGRESSION WITH APPLICATIONS TO UK BIOBANK.

In high-dimensional regression problems, often a relatively small subset of the features are relevant for predicting the outcome, and methods that impose sparsity on the solution are popular. When multiple correlated outcomes are available (multitask), reduced rank regression is an effective way to borrow strength and capture latent structures that underlie the data. Our proposal is motivated by the UK Biobank population-based cohort study, where we are faced with large-scale, ultrahigh-dimensional features, and have access to a large number of outcomes (phenotypes)-lifestyle measures, biomarkers, and disease outcomes. We are hence led to fit sparse reduced-rank regression models, using computational strategies that allow us to scale to problems of this size. We use a scheme that alternates between solving the sparse regression problem and solving the reduced rank decomposition. For the sparse regression component we propose a scalable iterative algorithm based on adaptive screening that leverages the sparsity assumption and enables us to focus on solving much smaller subproblems. The full solution is reconstructed and tested via an optimality condition to make sure it is a valid solution for the original problem. We further extend the method to cope with practical issues, such as the inclusion of confounding variables and imputation of missing values among the phenotypes. Experiments on both synthetic data and the UK Biobank data demonstrate the effectiveness of the method and the algorithm. We present multiSnpnet package, available at http://github.com/junyangq/multiSnpnet that works on top of PLINK2 files, which we anticipate to be a valuable tool for generating polygenic risk scores from human genetic studies.

Sparse reduced-rank regression for simultaneous rank and variable selection via manifold optimization

We consider the problem of constructing a reduced-rank regression model whose coefficient parameter is represented as a singular value decomposition with sparse singular vectors. The traditional estimation procedure for the coefficient parameter often fails when the true rank of the parameter is high. To overcome this issue, we develop an estimation algorithm with rank and variable selection via sparse regularization and manifold optimization, which enables us to obtain an accurate estimation of the coefficient parameter even if the true rank of the coefficient parameter is high. Using sparse regularization, we can also select an optimal value of the rank. We conduct Monte Carlo experiments and a real data analysis to illustrate the effectiveness of our proposed method.

Association of dietary inflammatory potential with risk of overall and cause-specific mortality.

Inflammation is a central mechanism in metabolic disorders associated with morbidity and mortality and dietary factors can modulate inflammation. We aimed to prospectively investigate the association between an empirically developed, food-based dietary inflammatory pattern (EDIP) score and the risk of overall and cause-specific mortality, using data from the US National Health and Nutrition Examination Survey from 1999 to 2014. EDIP score was derived by entering thirty-nine predefined commonly consumed food groups into the reduced rank regression models followed by stepwise linear regression, which was most predictive of two plasma inflammation biomarkers including C-reactive protein and leucocyte count among 25500 US adults. This score was further validated in a testing set of 9466 adults. Deaths from baseline until 31 December 2015 were identified through record linkage to the National Death Index. During a median follow-up of 7·8 years among 40074 participants, we documented 4904 deaths. Compared with participants in the lowest quintile of EDIP score, those in the highest quintile had a higher risk of overall death (hazard ratio (HR) = 1·19, 95 % CI 1·08, 1·32, Ptrend = 0·002), and deaths from cancer (HR = 1·41, 95 % CI 1·14, 1·74, Ptrend = 0·017) and CVD (HR = 1·22, 95 % CI 0·98, 1·53, Ptrend = 0·211). When stratified by age, the association of EDIP with overall mortality was stronger among individuals under 65 years of age (Pinteraction = 0·001). Diets with a higher inflammatory potential were associated with increased risk of overall and cancer-specific mortality. Interventions to reduce the adverse effect of pro-inflammatory diets may potentially promote health and longevity.

Development of Empirical Dietary Inflammatory Pattern (EDIP) scores with Different Food Groups and Biomarkers

Background: The empirical dietary inflammatory pattern (EDIP) is a hypothesis-driven dietary pattern used to assess the inflammatory potential of diet in the US population. Food-frequency questionnaire responses are used to build regression models comparing this dietary information to circulating inflammatory profiles, to help determine which food groups have more or less inflammatory potential on specific individuals. We will eventually use this tool in a cancer patient intervention to modify inflammation and improve chances of survival.&#x0D; Methods: EDIP scores were calculated for 4 models from 24hr recalls reported by 67 women noncancer controls that had signed an informed consent prior to participation. The Luminex Human Chemokine Multiplex Assay was used to measure 11 chemokines and cytokines. As seen in previous studies, we first derived a model, EDIP-Limited (EDIP-L), by using a reduced rank regression model of all 17 food groups followed by a multivariable regression analysis to identify a dietary pattern that predicts concentrations of two inflammatory biomarkers: IL-6 and TNF-a. We derived a secondary EDIP score using a new model, EDIP-All Inclusive (EDIP-AI), which included the same 17 food groups to predict all 11 circulating biomarkers in our panel. Lastly, we developed two additional EDIP models to test how the biomarker predication may change when we regrouped our food variables from 17 to 14 groupings. EDIP-Limited New (EDIP-LN) used 14 new food groups derived from the same 24hr recalls, only predicting IL-6 and TNF-a. EDIP-All New (EDIP-AN) used those same 14 food groups with all 11 biomarkers.&#x0D; Results: In this study, we optimize models for EDIP and report the differences in EDIP scores based on the inflammatory biomarkers and food groups used in analysis. Briefly, the components of EDIP-L were not significant. After including all the biomarkers, the components of EDIP-AI were: “fruit juice” (p = 0.0009), “snacks” (p = 0.0008), “leafy green vegetables” (p = 0.0074), “low-energy beverages” (p = 0.0098), “red meat” (p = 0.0038), “fruit” (p = 0.0002) and “whole wheat grains” (p = 0.0138). Similarly, after reorganizing our food items into 14 food groups, the components of EDIP-LN were not significant. However, components of EDIP-AN were: “fruit juice” (p = 0.0107), “snacks” (p = 0.0116) and “fruit” (p = 0.0026).&#x0D; Conclusions: Findings demonstrate the EDIP scores differ based on the inflammatory biomarkers and food groups used in the analysis on the same noncancer controls. Depending on the methods used, an individual’s diet may be considered more pro- or anti-inflammatory. This study provides insight into the inflammatory potential of an individual’s diet and the factors that may affect how we calculate this potential.

Alternating DCA for reduced-rank multitask linear regression with covariance matrix estimation

We study a challenging problem in machine learning that is the reduced-rank multitask linear regression with covariance matrix estimation. The objective is to build a linear relationship between multiple output variables and input variables of a multitask learning process, taking into account the general covariance structure for the errors of the regression model in one hand, and reduced-rank regression model in another hand. The problem is formulated as minimizing a nonconvex function in two joint matrix variables (X,Θ) under the low-rank constraint on X and positive definiteness constraint on Θ. It has a double difficulty due to the non-convexity of the objective function as well as the low-rank constraint. We investigate a nonconvex, nonsmooth optimization approach based on DC (Difference of Convex functions) programming and DCA (DC Algorithm) for this hard problem. A penalty reformulation is considered which takes the form of a partial DC program. An alternating DCA and its inexact version are developed, both algorithms converge to a weak critical point of the considered problem. Numerical experiments are performed on several synthetic and benchmark real multitask linear regression datasets. The numerical results show the performance of the proposed algorithms and their superiority compared with three classical alternating/joint methods.

An Explicit Mean-Covariance Parameterization for Multivariate Response Linear Regression

We develop a new method to fit the multivariate response linear regression model that exploits a parametric link between the regression coefficient matrix and the error covariance matrix. Specifically, we assume that the correlations between entries in the multivariate error random vector are proportional to the cosines of the angles between their corresponding regression coefficient matrix columns, so as the angle between two regression coefficient matrix columns decreases, the correlation between the corresponding errors increases. We highlight two models under which this parameterization arises: a latent variable reduced-rank regression model and the errors-in-variables regression model. We propose a novel nonconvex weighted residual sum of squares criterion which exploits this parameterization and admits a new class of penalized estimators. The optimization is solved with an accelerated proximal gradient descent algorithm. Our method is used to study the association between microRNA expression and cancer drug activity measured on the NCI-60 cell lines. An R package implementing our method, MCMVR, is available online.

Reduced Rank Regression Models in Economics and Finance

This chapter surveys the importance of reduced rank regression techniques (RRR) for modelling economic and ?nancial time series. We mainly focus on models that are capable to reproduce the presence of common dynamics among variables such as the serial correlation common feature and the multivariate autoregressive index models. Cointegration analysis, for which RRR plays a central role, is not discussed in this chapter as it deserves a speci?c treatment on its own. Instead, we show how to detect and model comovements in time series that are stationary or that have been stationarized after proper transformations. The motivations for the use of RRR in time series econometrics include dimension reductions which simplify complex dynamics and thus making interpretations easier, as well as pursuing e¢ ciency gains in both estimation and prediction. Via the ?nal equation representation, RRR also makes the nexus between multivariate time series and parsimonious marginal ARIMA models. The drawback of RRR, which is common to all the dimension reduction techniques, is that the underlying restrictions may be present or not in the data. We provide in this chapter a couple of empirical applications to illustrate concepts and methods.

Abstract PO-142: Development of Empirical Dietary Inflammatory Pattern (EDIP) scores with different food groups and biomarkers

Abstract Background: The empirical dietary inflammatory pattern (EDIP) is a hypothesis- driven dietary pattern used to assess the inflammatory potential of diet in the US population. It is usually composed of food groups obtained from a food-frequency questionnaire. Methods: EDIP scores were calculated for 4 models from 24hr recalls reported by 67 women noncancer controls that had signed an informed consent prior to participation. The Luminex Human Chemokine Multiplex Assay was used to measure 11 chemokines and cytokines. As seen in previous studies, we first derived a model, EDIP-Limited (EDIP-L), by entering 18 food groups in reduced rank regression models followed by a multivariable regression analysis to identify a dietary pattern that predicts concentrations of two inflammatory biomarkers: IL-6 and TNF-a. We derived another EDIP score using a new model, EDIP-All Inclusive (EDIP-AI), which included the same 18 food groups to predict all 11 biomarkers. Lastly, we developed two other EDIP models. EDIP-Limited New (EDIP-LN) used 14 new food groups derived from the same 24hr recalls, only predicting IL-6 and TNF-a. EDIP-All New (EDIP-AN) used those 14 food groups with all 11 biomarkers. Results: In this study, we optimize models for EDIP and report the differences in EDIP scores based on the inflammatory biomarkers and food groups used in analysis. Briefly, the components of EDIP-L were not significant. After including all the biomarkers, the components of EDIP-AI were: “fruit juice” (p = 0.0009), “snacks” (p = 0.0008), “leafy green vegetables” (p = 0.0074), “low-energy beverages” (p = 0.0098), “red meat” (p = 0.0038), “fruit” (p = 0.0002) and “whole wheat grains” (p = 0.0138). Similarly, after reorganizing our food items into 14 food groups, the components of EDIP-LN were not significant. However, components of EDIP-AN were: “fruit juice” (p = 0.0107), “snacks” (p = 0.0116) and “fruit” (p = 0.0026). Conclusions: Findings demonstrate the EDIP scores differ based on the inflammatory biomarkers and food groups used in the analysis on the same noncancer controls. Depending on the methods used, an individual’s diet may be considered more pro- or anti-inflammatory. This study provides insight into the inflammatory potential of an individual’s diet and the factors that may affect how we calculate this potential. Citation Format: Emma Guyonnet, Millicent Amankwah, Yalei Chen, Rachel Martini, Melissa Davis, Lisa Newman. Development of Empirical Dietary Inflammatory Pattern (EDIP) scores with different food groups and biomarkers [abstract]. In: Proceedings of the AACR Virtual Conference: Thirteenth AACR Conference on the Science of Cancer Health Disparities in Racial/Ethnic Minorities and the Medically Underserved; 2020 Oct 2-4. Philadelphia (PA): AACR; Cancer Epidemiol Biomarkers Prev 2020;29(12 Suppl):Abstract nr PO-142.