M-DATA: A statistical approach to jointly analyzing de novo mutations for multiple traits

Abstract

Recent studies have demonstrated that multiple early-onset diseases have shared risk genes, based on findings from de novo mutations (DNMs). Therefore, we may leverage information from one trait to improve statistical power to identify genes for another trait. However, there are few methods that can jointly analyze DNMs from multiple traits. In this study, we develop a framework called M-DATA (Multi-trait framework for De novo mutation Association Test with Annotations) to increase the statistical power of association analysis by integrating data from multiple correlated traits and their functional annotations. Using the number of DNMs from multiple diseases, we develop a method based on an Expectation-Maximization algorithm to both infer the degree of association between two diseases as well as to estimate the gene association probability for each disease. We apply our method to a case study of jointly analyzing data from congenital heart disease (CHD) and autism. Our method was able to identify 23 genes for CHD from joint analysis, including 12 novel genes, which is substantially more than single-trait analysis, leading to novel insights into CHD disease etiology.

Statistical methods for assessing the effects of de novo variants on birth defects

“…M-DATA is another multi-trait method that shares the same hypotheses as mTADA, but uses an Expectation-Maximization algorithm to estimate parameters and infer risk gene status [ 51 ]. Compared to mTADA, this method uses an alternative way to characterize the effects of variants in different functional groups by linking variant-level and gene-level functional annotations to the relative risk of de novo genotype in the model.…”

Network assisted analysis of de novo variants using protein-protein interaction information identified 46 candidate genes for congenital heart disease

“…For network information, we first downloaded STRING v11.0 with medium edge likelihood via interface from STRINGdb package in R and call this original network from STRING G 0 . We obtained the curated list of known human CHD genes from Jin el al [5] and expanded the gene list by including additional candidate genes (FDR<0.1) from the single-trait analysis in our previous work [18]. This gene list (258 genes) was set as seed genes for our network.…”

“…A hierarchical Bayes strategy was adopted for parameter estimation in TADA. Following this idea, a number of methods have been proposed to improve TADA, with some focusing on leveraging the shared genetic information in multiple correlated phenotypes, such as neurodevelopmental disorders and CHD [17,18], whereas others extend the method by integrating DNMs with other types of genetic variants and functional annotations [19][20][21][22]. Please note that, except for DECO [22], all these methods treat each gene individually and do not consider the interaction effects of genes.…”

“…As an illustrative example, we applied TADA de novo test [16] with the CHD DNV data curated in our previous work [18], and conducted a post-association analysis on the p-values returned from the test. After false discovery rate (FDR) adjustment of p-values, we identified 21 genes with FDR<0.1 among 18,856 genes tested, and found that the number of edges formed by the 21 genes (20 edges, blue line in Fig 1A) is much larger than the upper tail of the empirical distribution sampled from 21 randomly selected genes in the STRING V11.0 database (score threshold: 400) for 10,000 times (Fig 1A).…”

Roles of statistical modeling in characterizing the genetic basis of human diseases and traits

“…Alternative methods are needed to quantify concordance of association signals for rare and de novo variants between traits [24]. With shared genetics between traits, we can leverage this to better identify disease genes [25,26].…”