Non Technical Summary
The sabbatical project will enhance the technical/data analytic skills and capabilities of the investigator,to perform cutting edge microbiome research. She will master new and analytical techniques related to strain and viral analyses from a top microbiome research group (Dr. Huttenhower). Her sabbatical stay will be heavily focused on mastering new cutting-edge techniques related to virome and strain analyses, she will also deepen her skill in the analyses of bacterial data.

Classification

Knowledge Area (KA)	Subject of Investigation (SOI)	Field of Science (FOS)	Percent
724	6099	1010	100%

Knowledge Area
724 - Healthy Lifestyle;

Subject Of Investigation
6099 - People and communities, general/other;

Field Of Science
1010 - Nutrition and metabolism;

Goals / Objectives
Thirteen percent of American adults have diabetes, of which 90-95% is type 2 diabetes.Diabetes disproportionally impacts Latinos at rates higher than non-Latino whites. The gut microbiota, a large and diverse community of microbes, its combined genome, the gut microbiome, exists in symbiosis with the host, plays a central role in human immune systemand metabolism. The gut microbiome has been heavily in the etiology of diabetes and insulin sensitivity. Diabetes and metabolic syndrome have unique microbial signatures. Importantly, the microbiome is modifiable and has potential for diabetes prevention and treatment, as part of personalized, diet-based therapy.The Mediterranean diet (MedDiet) is characterized by high consumption of fruits, vegetables, olive oil and whole grain cereals and is one of the most studied dietary patterns globally. Adherence to the MedDiet has been associated, with reduced risk of diabetes. Diet, including the MedDiet, has been shown to have strong impact on the gut microbiota. Adherence to the MedDiet has been associated with increase microbial diversity, and increased abundance of several anti-inflammatory species such as Eubacterium Eligens and the Faecalibacterium prausnitzii[18]. While the relationship between gut bacteria, diabetes and diet has been relatively widely studied, comparatively little work exists on US Latinos, who are subject to disparities in health, diet and disease. Latinos unique diet high in processed foods and a substantially increased risk of diabetes compared to non-Latino whites.Objective1. To examine the inter-relationship between healthy dietary patterns, the composition and function of the gut microbiome, including bacterial strains and diabetes in an under-represented Latino cohort.Objective 2. To examine the association between the gut virome and diabetes in Latinos.

Project Methods
METHODSi) Omnibus testing will seek to establish whether overall community patterns of variation (Alpha and Beta diversity) are associated with the MedDiet pattern. The Gini-Simpson Index for Alpha Diversity and the Bray Curtis dissimilarity for Beta diversity as generalizable metrics, will be used. The above analyses will be performed using the R package Vegan(R).ii) feature-wise analyses, in MaAsLin2 to will identify taxa associated with MedDiet in the BPRHS.MaAsLin2 performs per-feature multivariate testing, adjusting for metadata covariates, and accounting for multiple hypothesis testing using FDR.iii) We will use Random forest (RF) classifier to evaluate how well differences in taxonomicprofiles perform in classifying BPHRS participants according to high (above the median) vs. low (below the median) adherence to MedDiet diet pattern. RF is microbiome-appropriate discriminative prediction model.In supplemental exploratory analyses, we will examine the association between bacterial strains and diabetes in Latinos. We will examine the role of diet in the bacterial strain-diabetes relationship.The Sabbatical applicat will master two key tools for strain analyses: AnPan and StrainPhlAn.StrainPhlAnis a novel metagenomic strain identification approach. The newest versions of MetaPhlAn2.8 and StrainPhlAn 2.8 include 5 times more genomes than the currently available public version. The applicant has not used StrainPhLAn before and will master this tool during the sabbatical.AnPan: is an integrated package of statistical methods appropriate for identifying microbial strains associated with an outcome. In the proposed project Dr. Palacios will 1) learn how to use AnPan by working through tutorials and other information available in the Huttenhower group 2) apply AnPan to identify microbial strains associated with strong adherence to the MedDiet diet pattern. While MaAsLin2 (SA1 above) can identify bacterial pathways associated with disease, if a microbial pathway is only present in strains associated with disease, this is more challenging to detect with MaAsLin2 alone. AnPan can be used to detect strains associated with disease. AnPan allows us to detect pathways that are associated with an outcome, while accounting for species abundance.Association between the gut virome and diabetes in Latinos, examining the role of diet:Viral profiling: Dr. Palacios will work with Dr Huttenhower and his group to develop a deep understanding of how the Baqlava tool is used to generate metagenomic viral profiles. Baqlava is novel, not yet published tool that generates viral taxonomic profiles from metagenomic fastq files. Dr. Palacios will learn how Baqlava works and how to apply it to future analyses.Data analyses: Once the viral profiles are generated, the analytic approach in SA2, will be analogous to that applied to bacterial taxa in SA1, but considering viral profiles in lieu of bacterial taxonomy. Viral taxonomy will be aggregated at the phage level, and unclassified viruses will be removed. Linear regression will be used to relate Viral Alpha Diversity within each participant, summarized using the Gini-Simpson Index), to diabetes after adjusting potential confounders. We will use Principal Coordinates Analyses visually and PERMANOVA to statistically examine whether Beta Diversity varies with diabetes. We will perform feature-wise analyses, in MaAsLin2 to identify viral taxa associated with diabetes. We will use RF classifier to examine how well differences in viral composition perform in classifying participants according to whether they have diabetes and identify metagenomic viruses that contribute most strongly to the classification, via feature importance. For viruses related to diabetes, we will examine the association with MedDiet score. Analyses examining the association between virome and MedDiet score will follow the same paradigm as for diabetes above: 1) omnibus virome testing with PCoA visualization and PERMANOVA; 2) feature-wise analyses with MaAsLin2 and 3) RF classifier.