Eigen: a Spectral Approach to the Integration of Functional Genomics Annotations for Both Coding and Noncoding Sequence Variants

Over the past few years, substantial effort has been put into the functional annotation of variation in human genome sequence. Indeed, for any genetic variant, whether protein coding or noncoding, a diverse set of functional annotations is available from projects such as Ensembl, ENCODE and Roadmap Epigenomics. Such annotations can play a critical role in identifying putatively causal variants among the abundant natural variation that occurs at a locus of interest. The main challenges in using these various annotations include their large numbers, and their diversity. In particular, it is not clear a priori which annotation is better at predicting functionally relevant variants. It is therefore desirable to integrate these different annotations into a single measure of functional importance for a variant. Here we develop an unsupervised approach to derive such a meta-score (Eigen), that, unlike most existing methods, is not based on any labelled training data. Furthermore, the proposed method produces estimates of predictive accuracy for each functional annotation score, and subsequently uses these estimates of accuracy to derive the aggregate functional score for variants of interest as a weighted linear combination of individual annotations. We show that the resulting meta-score has better discriminatory ability using disease associated and putatively benign variants from published studies (for both Mendelian and complex diseases) compared with the recently proposed CADD score. In particular, we show that the proposed meta-score outperforms the CADD score on noncoding variants from GWAS and eQTL studies, noncoding somatic mutations in the COSMIC database, and on de novo coding mutations in epilepsy and autism studies. Across varied scenarios, the Eigen score performs generally better than any single individual annotation, representing a powerful single functional score that can be incorporated in fine-mapping studies.