Tuesday, April 23, 2019

New Gene Prioritization Toolkit adds value to GWAS results

Genetic association data from genome-wide association studies (GWAS) are foundational for our understanding of type 2 diabetes and other complex diseases. But in order to apply these results to diagnosis, drug development, and treatment, we need to identify the effector genes that explain those genetic associations. This is rarely straightforward: most SNPs associated with disease are located outside of coding regions of the genome, so that their impact on genes is not obvious; and even a variant located in a protein-coding gene may actually affect a different gene. And to complicate things further, a variant that is strongly associated with disease may not have a direct impact on a gene, but may rather be "along for the ride" with a tightly linked causal variant.

Today we have released a prototype, experimental version of an interactive tool in the Type 2 Diabetes Knowledge Portal that can help bridge the gap between genetic association results and the effector genes that are directly involved in disease. We are aggregating additional data types—for example, transcriptional regulation, tissue specificity, curated biological annotations, and more—and integrating them using cutting-edge computational methods in order to mine insights from GWAS data. The new Gene Prioritization Toolkit presents these data types and results to help researchers evaluate candidate causal genes around a genetic association signal.

As a first step in developing this tool, we needed to find a way to store many different connections between variants, genes, tissues, phenotypes, and biological annotations. We decided to use a Neo4J graph database, which holds data nodes and their relationships with each other and can support complex, scientifically meaningful queries.

Neo4J graph showing variants on chromosome 8 that are associated with glycemic phenotypes. Orange circles represent variants; pink, p-values; blue, phenotypes; red, phenotype group; green and brown, variant annotations.

We have also created pipelines to apply computational methods to the genetic association data in the T2DKP. In brief, we are currently running:
  • MetaXcan, which integrates tissue-specific expression data from GTEx and genetic association data to predict the potential that a gene is causal for a phenotype in a given tissue;
  • DEPICT, which integrates multiple data sources including transcriptional co-regulation, Gene Ontology annotations,  model organism phenotypes, and more to predict membership of a gene in a pathway and the probability of its association with a given phenotype;
  • eCAVIAR and COLOC, two methods that quantify the probability that a variant is causal in both genetic association and eQTL studies.
We present the results of these methods in an interactive table on a new tab of the Gene page (see an example), "Genes in region". 

In addition to the results of the methods listed above, the table includes gene-level T2D associations generated by two types of burden test (Firth and SKAT) from an analysis of nearly 50,000 exome sequences by Jason Flannick and colleagues, as well as the phenotypes of knockout mice that are mutant for homologs of the human genes in the region, from the Mouse Genome Database. All of these methods and data types are described in more detail in our downloadable help documentation for the new interface.

The table shows all of these data types for each gene across the region. It has two alternative views: the Significance view, in which table cells are color-coded by significance, and the Records view, in which shading indicates the number of records in each cell. This visual summary allows you to compare genes quickly across methods. Clicking on a cell opens a window listing full details of the results.

The table also supports versatile sorting. Columns may be dragged and dropped in order to group comparable genes, as shown below:

Default view of the Gene Prioritization table. Columns represent genes and rows represent methods or data types. Cell color denotes significance, with darker shades indicating higher significance.

The same table after custom re-ordering of columns to group three genes that all have significant eCAVIAR and COLOC scores.

In addition, the table may be transposed so that the columns represent methods and the rows represent genes. This allows sorting by significance within a method, so that the gene with the most significant result for each method is easily identified.

This entire system, from data storage through the computational pipelines through the user interface, has been designed to be flexible and modular so that in the future we will be able to add new methods and data types easily and rapidly. As we actively develop the system, we are very interested in feedback from researchers about how to improve it. Please try it out and let us know what you think!

Thursday, April 18, 2019

GPS information for BMI and obesity now available in the CVDKP

Genome-wide polygenic scores (GPS) have great potential for helping to advance research on complex diseases and traits. Not only can they help predict individual genetic risk, but they can also help us understand the physiology of disease, by identifying groups at the extremes of risk whose clinical profiles can be studied or who may be enrolled in clinical trials.

Following up on their previous work that generated GPSs for five complex diseases, co-lead authors Amit Khera and Mark Chaffin, along with senior author Sekar Kathiresan and colleagues, have now developed a GPS for body mass index (BMI) and obesity, published today in Cell. To help promote obesity research, the authors have provided an open-access file listing the variants and weights that comprise the GPS. That file is now available for download from the Downloads page of our sister Knowledge Portal, the Cardiovascular Disease Knowledge Portal.

To generate this GPS, Khera and colleagues started with a large, recently published genome-wide association study (GWAS) for BMI in more than 300,000 UK Biobank participants (Locke et al., 2015) and applied an algorithm that assigned a weight to each of 2.1 million variants, also taking into account factors such as the proportion of variants with non-zero effect size and the degree of correlation between a variant and its neighbors. They validated the GPS by applying it to nearly 120,000 additional UK Biobank participants, finding that the score was strongly correlated with measured BMI, and then applied it to four independent testing datasets.

We don't have space here to cover the many interesting details uncovered by the researchers, but overall, this work shows that a high GPS strongly predicts increased risk of severe obesity, cardiometabolic disease, and all-cause mortality. Those with the very highest GPS had a level of risk for obesity similar to that conferred by a rare monogenic mutation in the MC4R gene.

The GPS has the potential to be a powerful tool for people struggling with overweight and obesity. "Importantly, we are in the early days of identifying how we can best inform and empower patients to overcome health risks in their genetic background," said Khera in a press release from the Broad Institute. "We are incredibly excited about the potential to improve health outcomes."

We invite you to read the paper, take a look at the file of variants and weights freely available from the CVDKP Downloads page, and contact us with any questions!

Wednesday, April 3, 2019

New Hoorn DCS dataset available in the T2DKP via federation

A new dataset, "Hoorn DCS 2019," is now available in the Type 2 Diabetes Knowledge Portal via the T2DKP Federated node at the European Bioinformatics Institute (EBI). The Hoorn Diabetes Care System (DCS) cohort is a prospective cohort of type 2 diabetics in the West Friesland region of the Netherlands, for whom clinical measurements are collected annually. Association analysis was performed at EBI across 1,997 samples for 16 phenotypes, including glycemic, anthropometric, cardiovascular, and renal traits. The Hoorn DCS 2019 dataset is described in detail on the T2DKP Data page.

This new dataset is housed at the EBI Federated node of the T2DKP, which enables researchers to interact with results that may not be transferred to the AMP T2D Data Coordinating Center (DCC) at the Broad Institute because of institutional, regional, or national regulations. Data at the EBI node are stored in such a way that their specific privacy requirements are met, but they are available for secure remote queries via T2DKP tools and interfaces. Results from such queries are served up alongside results from all of the datasets housed at the AMP T2D DCC, such that researchers may browse and query data from any location without even needing to know where the data reside. This federation mechanism represents both an important technical advance in handling and protecting data, and a significant step forward in democratizing and improving access to genetic association results. Results at the EBI Federated node now comprise 9 datasets, nearly 40,000 samples, and associations for a wide variety of phenotypes.

Summary results from all of these datasets are integrated into Gene and Variant pages in the T2DKP, and may also be viewed in interactive Manhattan plots or queried using the Variant Finder tool. The individual-level data behind the datasets are accessible for custom association analysis in our Genetic Association Interactive Tool (GAIT) on Variant pages. Using this tool, researchers can filter samples to create a custom subset with defined characteristics such as age, gender, BMI, and other measures, and then run on-the-fly association analysis within that sample subset.

Please take a look at the new dataset and contact us with any questions or comments!