Predicting EGFR mutational status from pathology images using a real-world dataset.

Treatment of non-small cell lung cancer is increasingly biomarker driven with multiple genomic alterations, including those in the epidermal growth factor receptor (EGFR) gene, that benefit from targeted therapies. We developed a set of algorithms to assess EGFR status and morphology using a real-world advanced lung adenocarcinoma cohort of 2099 patients with hematoxylin and eosin (H&E) images exhibiting high morphological diversity and low tumor content relative to public datasets. The best performing EGFR algorithm was attention-based and achieved an area under the curve (AUC) of 0.870, a negative predictive value (NPV) of 0.954 and a positive predictive value (PPV) of 0.410 in a validation cohort reflecting the 15% prevalence of EGFR mutations in lung adenocarcinoma. The attention model outperformed a heuristic-based model focused exclusively on tumor regions, and we show that although the attention model also extracts signal primarily from tumor morphology, it extracts additional signal from non-tumor tissue regions. Further analysis of high-attention regions by pathologists showed associations of predicted EGFR negativity with solid growth patterns and higher peritumoral immune presence. This algorithm highlights the potential of deep learning tools to provide instantaneous rule-out screening for biomarker alterations and may help prioritize the use of scarce tissue for biomarker testing.

The Data Use Ontology to streamline responsible access to human biomedical datasets.

Human biomedical datasets that are critical for research and clinical studies to benefit human health also often contain sensitive or potentially identifying information of individual participants. Thus, care must be taken when they are processed and made available to comply with ethical and regulatory frameworks and informed consent data conditions. To enable and streamline data access for these biomedical datasets, the Global Alliance for Genomics and Health (GA4GH) Data Use and Researcher Identities (DURI) work stream developed and approved the Data Use Ontology (DUO) standard. DUO is a hierarchical vocabulary of human and machine-readable data use terms that consistently and unambiguously represents a dataset's allowable data uses. DUO has been implemented by major international stakeholders such as the Broad and Sanger Institutes and is currently used in annotation of over 200,000 datasets worldwide. Using DUO in data management and access facilitates researchers' discovery and access of relevant datasets. DUO annotations increase the FAIRness of datasets and support data linkages using common data use profiles when integrating the data for secondary analyses. DUO is implemented in the Web Ontology Language (OWL) and, to increase community awareness and engagement, hosted in an open, centralized GitHub repository. DUO, together with the GA4GH Passport standard, offers a new, efficient, and streamlined data authorization and access framework that has enabled increased sharing of biomedical datasets worldwide.

GA4GH: International policies and standards for data sharing across genomic research and healthcare.

The Global Alliance for Genomics and Health (GA4GH) aims to accelerate biomedical advances by enabling the responsible sharing of clinical and genomic data through both harmonized data aggregation and federated approaches. The decreasing cost of genomic sequencing (along with other genome-wide molecular assays) and increasing evidence of its clinical utility will soon drive the generation of sequence data from tens of millions of humans, with increasing levels of diversity. In this perspective, we present the GA4GH strategies for addressing the major challenges of this data revolution. We describe the GA4GH organization, which is fueled by the development efforts of eight Work Streams and informed by the needs of 24 Driver Projects and other key stakeholders. We present the GA4GH suite of secure, interoperable technical standards and policy frameworks and review the current status of standards, their relevance to key domains of research and clinical care, and future plans of GA4GH. Broad international participation in building, adopting, and deploying GA4GH standards and frameworks will catalyze an unprecedented effort in data sharing that will be critical to advancing genomic medicine and ensuring that all populations can access its benefits.

Empirical validation of an automated approach to data use oversight.

The current paradigm for data use oversight of biomedical datasets is onerous, extending the timescale and resources needed to obtain access for secondary analyses, thus hindering scientific discovery. For a researcher to utilize a controlled-access dataset, a data access committee must review her research plans to determine whether they are consistent with the data use limitations (DULs) specified by the informed consent form. The newly created GA4GH data use ontology (DUO) holds the potential to streamline this process by making data use oversight computable. Here, we describe an open-source software platform, the Data Use Oversight System (DUOS), that connects with DUO terminology to enable automated data use oversight. We analyze dbGaP data acquired since 2006, finding an exponential increase in data access requests, which will not be sustainable with current manual oversight review. We perform an empirical evaluation of DUOS and DUO on selected datasets from the Broad Institute's data repository. We were able to structure 118/123 of the evaluated DULs (96%) and 52/52 (100%) of research proposals using DUO terminology, and we find that DUOS' automated data access adjudication in all cases agreed with the DAC manual review. This first empirical evaluation of the feasibility of automated data use oversight demonstrates comparable accuracy to human-based data access oversight in real-world data governance.

Responsible sharing of biomedical data and biospecimens via the "Automatable Discovery and Access Matrix" (ADA-M).

Given the data-rich nature of modern biomedical research, there is a pressing need for a systematic, structured, computer-readable way to capture, communicate, and manage sharing rules that apply to biomedical resources. This is essential for responsible recording, versioning, communication, querying, and actioning of resource sharing plans. However, lack of a common "information model" for rules and conditions that govern the sharing of materials, methods, software, data, and knowledge creates a fundamental barrier. Without this, it can be virtually impossible for Research Ethics Committees (RECs), Institutional Review Boards (IRBs), Data Access Committees (DACs), biobanks, and end users to confidently track, manage, and interpret applicable legal and ethical requirements. This raises costs and burdens of data stewardship and decreases efficient and responsible access to data, biospecimens, and other resources. To address this, the GA4GH and IRDiRC organizations sponsored the creation of the Automatable Discovery and Access Matrix (ADA-M, read simply as "Adam"). ADA-M is a comprehensive information model that provides the basis for producing structured metadata "Profiles" of regulatory conditions, thereby enabling efficient application of those conditions across regulatory spheres. Widespread use of ADA-M will aid researchers in globally searching and prescreening potential data and/or biospecimen resources for compatibility with their research plans in a responsible and efficient manner, increasing likelihood of timely DAC approvals while also significantly reducing time and effort DACs, RECs, and IRBs spend evaluating resource requests and research proposals. Extensive online documentation, software support, video guides, and an Application Programming Interface (API) for ADA-M have been made available.

Registered access: authorizing data access.

The Global Alliance for Genomics and Health (GA4GH) proposes a data access policy model-"registered access"-to increase and improve access to data requiring an agreement to basic terms and conditions, such as the use of DNA sequence and health data in research. A registered access policy would enable a range of categories of users to gain access, starting with researchers and clinical care professionals. It would also facilitate general use and reuse of data but within the bounds of consent restrictions and other ethical obligations. In piloting registered access with the Scientific Demonstration data sharing projects of GA4GH, we provide additional ethics, policy and technical guidance to facilitate the implementation of this access model in an international setting.

Simplifying research access to genomics and health data with Library Cards.

The volume of genomics and health data is growing rapidly, driven by sequencing for both research and clinical use. However, under current practices, the data is fragmented into many distinct datasets, and researchers must go through a separate application process for each dataset. This is time-consuming both for the researchers and the data stewards, and it reduces the velocity of research and new discoveries that could improve human health. We propose to simplify this process, by introducing a standard Library Card that identifies and authenticates researchers across all participating datasets. Each researcher would only need to apply once to establish their bona fides as a qualified researcher, and could then use the Library Card to access a wide range of datasets that use a compatible data access policy and authentication protocol.

High-throughput identification of RNA nuclear enrichment sequences.

In the post-genomic era, thousands of putative noncoding regulatory regions have been identified, such as enhancers, promoters, long noncoding RNAs (lncRNAs), and a cadre of small peptides. These ever-growing catalogs require high-throughput assays to test their functionality at scale. Massively parallel reporter assays have greatly enhanced the understanding of noncoding DNA elements en masse Here, we present a massively parallel RNA assay (MPRNA) that can assay 10,000 or more RNA segments for RNA-based functionality. We applied MPRNA to identify RNA-based nuclear localization domains harbored in lncRNAs. We examined a pool of 11,969 oligos densely tiling 38 human lncRNAs that were fused to a cytosolic transcript. After cell fractionation and barcode sequencing, we identified 109 unique RNA regions that significantly enriched this cytosolic transcript in the nucleus including a cytosine-rich motif. These nuclear enrichment sequences are highly conserved and over-represented in global nuclear fractionation sequencing. Importantly, many of these regions were independently validated by single-molecule RNA fluorescence in situ hybridization. Overall, we demonstrate the utility of MPRNA for future investigation of RNA-based functionalities.

Visualization of lncRNA by single-molecule fluorescence in situ hybridization.

Single-molecule RNA fluorescence in situ hybridization is a technique that holds great potential for the study of long noncoding RNA. It enables quantification and spatial resolution of single RNA molecules within cells via hybridization of multiple, labeled nucleic acid probes to a target RNA. It has recently become apparent that single-molecule RNA FISH probes targeting noncoding RNA are more prone to off-target binding yielding spurious results than when targeting mRNA. Here we present a protocol for the application of single-molecule RNA FISH to the study of noncoding RNA as well as an experimental procedure for validating legitimate signals.

Localization and abundance analysis of human lncRNAs at single-cell and single-molecule resolution.

BACKGROUND: Long non-coding RNAs (lncRNAs) have been implicated in diverse biological processes. In contrast to extensive genomic annotation of lncRNA transcripts, far fewer have been characterized for subcellular localization and cell-to-cell variability. Addressing this requires systematic, direct visualization of lncRNAs in single cells at single-molecule resolution. RESULTS: We use single-molecule RNA-FISH to systematically quantify and categorize the subcellular localization patterns of a representative set of 61 lncRNAs in three different cell types. Our survey yields high-resolution quantification and stringent validation of the number and spatial positions of these lncRNA, with an mRNA set for comparison. Using this highly quantitative image-based dataset, we observe a variety of subcellular localization patterns, ranging from bright sub-nuclear foci to almost exclusively cytoplasmic localization. We also find that the low abundance of lncRNAs observed from cell population measurements cannot be explained by high expression in a small subset of 'jackpot' cells. Additionally, nuclear lncRNA foci dissolve during mitosis and become widely dispersed, suggesting these lncRNAs are not mitotic bookmarking factors. Moreover, we see that divergently transcribed lncRNAs do not always correlate with their cognate mRNA, nor do they have a characteristic localization pattern. CONCLUSIONS: Our systematic, high-resolution survey of lncRNA localization reveals aspects of lncRNAs that are similar to mRNAs, such as cell-to-cell variability, but also several distinct properties. These characteristics may correspond to particular functional roles. Our study also provides a quantitative description of lncRNAs at the single-cell level and a universally applicable framework for future study and validation of lncRNAs.

linc-HOXA1 is a noncoding RNA that represses Hoxa1 transcription in cis.

Recently, researchers have uncovered the presence of many long noncoding RNAs (lncRNAs) in embryonic stem cells and believe they are important regulators of the differentiation process. However, there are only a few examples explicitly linking lncRNA activity to transcriptional regulation. Here, we used transcript counting and spatial localization to characterize a lncRNA (dubbed linc-HOXA1) located ∼50 kb from the Hoxa gene cluster in mouse embryonic stem cells. Single-cell transcript counting revealed that linc-HOXA1 and Hoxa1 RNA are highly variable at the single-cell level and that whenever linc-HOXA1 RNA abundance was high, Hoxa1 mRNA abundance was low and vice versa. Knockdown analysis revealed that depletion of linc-HOXA1 RNA at its site of transcription increased transcription of the Hoxa1 gene cis to the chromosome and that exposure of cells to retinoic acid can disrupt this interaction. We further showed that linc-HOXA1 RNA represses Hoxa1 by recruiting the protein PURB as a transcriptional cofactor. Our results highlight the power of transcript visualization to characterize lncRNA function and also suggest that PURB can facilitate lncRNA-mediated transcriptional regulation.

Mutations causing medullary cystic kidney disease type 1 lie in a large VNTR in MUC1 missed by massively parallel sequencing.

Although genetic lesions responsible for some mendelian disorders can be rapidly discovered through massively parallel sequencing of whole genomes or exomes, not all diseases readily yield to such efforts. We describe the illustrative case of the simple mendelian disorder medullary cystic kidney disease type 1 (MCKD1), mapped more than a decade ago to a 2-Mb region on chromosome 1. Ultimately, only by cloning, capillary sequencing and de novo assembly did we find that each of six families with MCKD1 harbors an equivalent but apparently independently arising mutation in sequence markedly under-represented in massively parallel sequencing data: the insertion of a single cytosine in one copy (but a different copy in each family) of the repeat unit comprising the extremely long (∼1.5-5 kb), GC-rich (>80%) coding variable-number tandem repeat (VNTR) sequence in the MUC1 gene encoding mucin 1. These results provide a cautionary tale about the challenges in identifying the genes responsible for mendelian, let alone more complex, disorders through massively parallel sequencing.

Peptidomic discovery of short open reading frame-encoded peptides in human cells.

The complete extent to which the human genome is translated into polypeptides is of fundamental importance. We report a peptidomic strategy to detect short open reading frame (sORF)-encoded polypeptides (SEPs) in human cells. We identify 90 SEPs, 86 of which are previously uncharacterized, which is the largest number of human SEPs ever reported. SEP abundances range from 10-1,000 molecules per cell, identical to abundances of known proteins. SEPs arise from sORFs in noncoding RNAs as well as multicistronic mRNAs, and many SEPs initiate with non-AUG start codons, indicating that noncanonical translation may be more widespread in mammals than previously thought. In addition, coding sORFs are present in a small fraction (8 out of 1,866) of long intergenic noncoding RNAs. Together, these results provide strong evidence that the human proteome is more complex than previously appreciated.

A Keystone for ncRNA.

A report on the Keystone symposium 'Non-coding RNAs' held at Snowbird, Utah, USA, 31 March to 5 April 2012.

Integrative annotation of human large intergenic noncoding RNAs reveals global properties and specific subclasses.

Large intergenic noncoding RNAs (lincRNAs) are emerging as key regulators of diverse cellular processes. Determining the function of individual lincRNAs remains a challenge. Recent advances in RNA sequencing (RNA-seq) and computational methods allow for an unprecedented analysis of such transcripts. Here, we present an integrative approach to define a reference catalog of >8000 human lincRNAs. Our catalog unifies previously existing annotation sources with transcripts we assembled from RNA-seq data collected from ∼4 billion RNA-seq reads across 24 tissues and cell types. We characterize each lincRNA by a panorama of >30 properties, including sequence, structural, transcriptional, and orthology features. We found that lincRNA expression is strikingly tissue-specific compared with coding genes, and that lincRNAs are typically coexpressed with their neighboring genes, albeit to an extent similar to that of pairs of neighboring protein-coding genes. We distinguish an additional subset of transcripts that have high evolutionary conservation but may include short ORFs and may serve as either lincRNAs or small peptides. Our integrated, comprehensive, yet conservative reference catalog of human lincRNAs reveals the global properties of lincRNAs and will facilitate experimental studies and further functional classification of these genes.

Large intergenic non-coding RNA-RoR modulates reprogramming of human induced pluripotent stem cells.

The conversion of lineage-committed cells to induced pluripotent stem cells (iPSCs) by reprogramming is accompanied by a global remodeling of the epigenome, resulting in altered patterns of gene expression. Here we characterize the transcriptional reorganization of large intergenic non-coding RNAs (lincRNAs) that occurs upon derivation of human iPSCs and identify numerous lincRNAs whose expression is linked to pluripotency. Among these, we defined ten lincRNAs whose expression was elevated in iPSCs compared with embryonic stem cells, suggesting that their activation may promote the emergence of iPSCs. Supporting this, our results indicate that these lincRNAs are direct targets of key pluripotency transcription factors. Using loss-of-function and gain-of-function approaches, we found that one such lincRNA (lincRNA-RoR) modulates reprogramming, thus providing a first demonstration for critical functions of lincRNAs in the derivation of pluripotent stem cells.

Predicting metabolic biomarkers of human inborn errors of metabolism.

Early diagnosis of inborn errors of metabolism is commonly performed through biofluid metabolomics, which detects specific metabolic biomarkers whose concentration is altered due to genomic mutations. The identification of new biomarkers is of major importance to biomedical research and is usually performed through data mining of metabolomic data. After the recent publication of the genome-scale network model of human metabolism, we present a novel computational approach for systematically predicting metabolic biomarkers in stochiometric metabolic models. Applying the method to predict biomarkers for disruptions of red-blood cell metabolism demonstrates a marked correlation with altered metabolic concentrations inferred through kinetic model simulations. Applying the method to the genome-scale human model reveals a set of 233 metabolites whose concentration is predicted to be either elevated or reduced as a result of 176 possible dysfunctional enzymes. The method's predictions are shown to significantly correlate with known disease biomarkers and to predict many novel potential biomarkers. Using this method to prioritize metabolite measurement experiments to identify new biomarkers can provide an order of a 10-fold increase in biomarker detection performance.

Chromatin signature reveals over a thousand highly conserved large non-coding RNAs in mammals.

There is growing recognition that mammalian cells produce many thousands of large intergenic transcripts. However, the functional significance of these transcripts has been particularly controversial. Although there are some well-characterized examples, most (>95%) show little evidence of evolutionary conservation and have been suggested to represent transcriptional noise. Here we report a new approach to identifying large non-coding RNAs using chromatin-state maps to discover discrete transcriptional units intervening known protein-coding loci. Our approach identified approximately 1,600 large multi-exonic RNAs across four mouse cell types. In sharp contrast to previous collections, these large intervening non-coding RNAs (lincRNAs) show strong purifying selection in their genomic loci, exonic sequences and promoter regions, with greater than 95% showing clear evolutionary conservation. We also developed a functional genomics approach that assigns putative functions to each lincRNA, demonstrating a diverse range of roles for lincRNAs in processes from embryonic stem cell pluripotency to cell proliferation. We obtained independent functional validation for the predictions for over 100 lincRNAs, using cell-based assays. In particular, we demonstrate that specific lincRNAs are transcriptionally regulated by key transcription factors in these processes such as p53, NFkappaB, Sox2, Oct4 (also known as Pou5f1) and Nanog. Together, these results define a unique collection of functional lincRNAs that are highly conserved and implicated in diverse biological processes.

Network-based prediction of human tissue-specific metabolism.

Direct in vivo investigation of mammalian metabolism is complicated by the distinct metabolic functions of different tissues. We present a computational method that successfully describes the tissue specificity of human metabolism on a large scale. By integrating tissue-specific gene- and protein-expression data with an existing comprehensive reconstruction of the global human metabolic network, we predict tissue-specific metabolic activity in ten human tissues. This reveals a central role for post-transcriptional regulation in shaping tissue-specific metabolic activity profiles. The predicted tissue specificity of genes responsible for metabolic diseases and tissue-specific differences in metabolite exchange with biofluids extend markedly beyond tissue-specific differences manifest in enzyme-expression data, and are validated by large-scale mining of tissue-specificity data. Our results establish a computational basis for the genome-wide study of normal and abnormal human metabolism in a tissue-specific manner.