Model, Integrate, Search... Repeat: A Sound Approach to Building Integrated Repositories of Genomic Data

A wealth of public data repositories is available to drive genomics and clinical research. However, there is no agreement among the various data formats and models; in the common practice, data sources are accessed one by one, learning their specific descriptions with tedious efforts. In this context, the integration of genomic data and of their describing metadata becomes—at the same time—an important, difficult, and well-recognized challenge. In this chapter, after overviewing the most important human genomic data players, we propose a conceptual model of metadata and an extended architecture for integrating datasets, retrieved from a variety of data sources, based upon a structured transformation process; we then describe a user-friendly search system providing access to the resulting consolidated repository, enriched by a multi-ontology knowledge base. Inspired by our work on genomic data integration, during the COVID-19 pandemic outbreak we successfully re-applied the previously proposed model-build-search paradigm, building on the analogies among the human and viral genomics domains. The availability of conceptual models, related databases, and search systems for both humans and viruses will provide important opportunities for research, especially if virus data will be connected to its host, provider of genomic and phenotype information.

Total Functional Score of Enhancer Elements Identifies Lineage-Specific Enhancers That Drive Differentiation of Pancreatic Cells

The differentiation of embryonic stem cells into various lineages is highly dependent on the chromatin state of the genome and patterns of gene expression. To identify lineage-specific enhancers driving the differentiation of progenitors into pancreatic cells, we used a previously described computational framework called Total Functional Score of Enhancer Elements (TFSEE), which integrates multiple genomic assays that probe both transcriptional and epigenomic states. First, we evaluated and compared TFSEE as an enhancer-calling algorithm with enhancers called using GRO-seq-defined enhancer transcripts (method 1) versus enhancers called using histone modification ChIP-seq data (method 2). Second, we used TFSEE to define the enhancer landscape and identify transcription factors (TFs) that maintain the multipotency of a subpopulation of endodermal stem cells during differentiation into pancreatic lineages. Collectively, our results demonstrate that TFSEE is a robust enhancer-calling algorithm that can be used to perform multilayer genomic data integration to uncover cell type-specific TFs that control lineage-specific enhancers.

Best Paper Selection

Lee SI, Celik S, Logsdon BA, Lundberg SM, Martins TJ, Oehler VG, Estey EH, Miller CP, Chien S, Dai J, Saxena A, Blau CA, Becker PS. A machine learning approach to integrate big data for precision medicine in acute myeloid leukemia. Nat Commun 2018 Jan;9(1):42 https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5752671/ Mobadersany P, Yousefi S, Amgad M, Gutman DA, Barnholtz-Sloan JS, Velázquez Vega JE, Brat DJ, Cooper LAD. Predicting cancer outcomes from histology and genomics using convolutional networks. Proc Natl Acad Sci U S A 2018;115(13):E2970-E2979 https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5879673/ Sengupta S, Sun SQ, Huang KL, Oh C, Bailey MH, Varghese R, Wyczalkowski MA, Ning J, Tripathi P, Mc Michael JF, Johnson KJ, Kandoth C, Welch J, Ma C, Wendl MC, Payne SH, Fenyö D, Townsend RR, Dipersio JF, Chen F, Ding L. Integrative omics analyses broaden treatment targets in human cancer. Genome Med 2018 Jul 27;10(1):60 https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6064051/ Torshizi AD, Petzold LR. Graph-based semi-supervised learning with genomic data integration using condition-responsive genes applied to phenotype classification. J Am Med Inform Assoc 2018;25(1):99-108 https://academic.oup.com/jamia/article/25/1/99/3826530