4 History of the Encyclopedia of DNA Elements (ENCODE) Project

1.1.1. Pilot Project

Based on peer review, programmatic and scientific balance, and input from the NACHGR, NHGRI funded eight awards for pilot projects (National Human Genome Research Institute 2012b). Although all funded awards were made to researchers outside of the NIH, several NHGRI intramural groups were actively involved in the ENCODE Pilot Project, most notably one led by Eric Green. His group’s role was to select homologous bacterial artificial chromosome clones from DNA libraries constructed from the DNA of a wide range of non-human species and sequence them to identify highly conserved regions that were expected to be functionally important by virtue of being evolutionarily conserved.

While the methods and technologies used in the different phases of ENCODE evolved, the basic premise of most of the experimental projects was to apply high-throughput assays to measure biochemical activities associated with specific regions of the genome that had previously been mechanistically associated with particular classes of functional genomic elements (see Figure 4.1). For example, projects mapped binding sites of transcription factors and histone modifications that were associated with promoters and enhancers. However, it is important to point out that this approach identifies genomic regions that have the potential to have functional activity, and additional work is needed to determine if, when, and where these regions have functions in particular biological settings (see below). Unlike the genome sequence, which is nearly identical in every cell in an organism, most genes and non-coding sequences that regulate the genome are active only in specific biological contexts.

Several key steps were needed to get the Consortium launched. An organizing committee was convened prior to the funding of the Pilot Project awards. One of its main tasks was to select target sequences comprising the 1 percent (~30Mb) to be analyzed. The committee decided on a mixed strategy. Half of the sequences would be selected manually based on the existence of substantial prior knowledge about genes and/or other functional regions and for which substantial comparative genomic data was available. Fourteen target regions ranging in size from 0.5 to 2 Mb (totaling ~15Mb) were selected. The other half of the ~30Mb were thirty 0.5Mb regions that were selected using a “stratified random-sampling strategy based on gene density and level of non-exonic conservation” to include regions with a range of gene and functional element density (National Human Genome Research Institute 2012a).

As described in more detail below, tight coordination and project management enabled by frequent communication had been an essential feature in the success of the HGP and ENCODE followed this successful model. For example, monthly ENCODE teleconference calls were held, fittingly, on Fridays at 11:00 a.m. ET, which previously had been the timeslot for the HGP sequencing teleconference calls. This time of day was chosen to accommodate participants across a wide range of time zones, including GMT (4:00 p.m.) and PT (8:00 a.m.). Also, discussed in more detail below, was the early recognition for the need to have a centralized repository to store, display, and make data available.

Since ENCODE was a new project, the Consortium decided to write a “Marker Paper”—a brief introduction to the project including a description of its goals—that was published in Science in October 2004 (ENCODE Consortium 2004). It laid out the research strategy, including target sequences, organization, operating principles, plans for data management, and long-term goals. The writing of a Marker Paper was in keeping with NHGRI’s practice of informing the community about what resources they could anticipate having access to and striving for transparency and inclusion (Strausberg et al. 1999; ENCODE Consortium 2004).

The pilot project concluded with series of papers including a capstone publication (Birney et al. 2007) and twenty-eight companion papers in the Genome Research ENCODE Special Issue (ENCODE Consortium 2007b). As the first major set of publications from ENCODE, this served to bring the community’s attention to the project at a time when the next phase of ENCODE was being planned and modENCODE (see below) was just being initiated. The process of writing and publishing a Consortium-wide paper along with companion papers, however, was met with a number of challenges that presaged challenges in future ENCODE and modENCODE publications (see below). For example, the Consortium initially submitted five independent papers on different themes, including transcription and gene regulation that were of varying strength. Based on the reviewers’ and editors’ feedback, these five papers were condensed into the capstone paper.

Some of the key findings and surprises included the discovery that most of the ~30Mb of the human genome studied in the pilot is transcribed into RNA, that about half of the potential functional elements identified do not appear to be located in evolutionarily conserved regions, and that regulatory elements are found within and downstream of genes more frequently than previously thought. Most importantly, the pilot project demonstrated the feasibility of its approach to functionally annotate the human genome.

1.1.2. ENCODE 2

Based on the success of the pilot project, NHGRI decided to fund another phase of ENCODE (ENCODE 2) that continued support of pilot projects focused on the selected target regions as well as supported new projects that scaled to whole-genome analysis. The plan was endorsed by the NACHGR in May 2006 (National Human Genome Research Institute 2006d), and three RFAs were subsequently released and applications funded (National Human Genome Research Institute 2018a): (1) To support data production: Creating the Encyclopedia of DNA Elements (ENCODE) in the Human Genome (U01 and U54) (National Human Genome Research Institute 2006a); (2) To support data management: A Data Coordination Center for the Encyclopedia of DNA Elements (ENCODE) Project (U41) (National Human Genome Research Institute 2006b); and (3) To support data analysis: A Data Analysis Center for the Encyclopedia of DNA Elements (ENCODE) Project (U01) (National Human Genome Research Institute 2007a). Two additional RFAs were funded to support additional technology development.

Major technological advances in DNA sequencing at this time enabled the adaptation of previously used assays reliant on microarray readouts to sequencing readouts as well as the development of new sequence-based approaches to map biochemical activities; the advances increased the feasibility of whole-genome analyses. (See Technology Development section below.) However, in these early days of implementing the new technologies, uncertainties for success in making the leap from 1 percent to the entire genome were of concern. To mitigate these uncertainties, NHGRI supported several groups working on similar approaches and made provisions to move funds between groups in the later years if the need arose, although in the end that was not necessary.

Many of the observations that had been made in the pilot project were confirmed and expanded upon in this second phase of ENCODE. ENCODE 2 provided a richer and more complex view of the human genome, identifying a vast number of sequences with potential regulatory activity that control when and where genes are expressed. For example, experimental evidence was found that regulatory elements can work from a considerable distance more often than had been previously realized. Many disease-associated genetic variants were found to be located within the non-coding regulatory regions that ENCODE identified, which was surprising at the time, as it was expected that most would be located within protein-coding regions and amplified the importance of studying non-coding sequences in a systematic way. By studying more than one hundred different cell types, additional observations could be made about which functional elements are expressed in which cell types, enabling the study of understanding why different types of cells have different characteristics and functions—and this only increased as ENCODE expanded the repertoire of cell types studied in ensuing years. What became evident was the power of genome-wide functional data to enable the derivation of generalizable observations on genome structure and function that heretofore was based on a limited number of genes or genomic regions.

In September 2012, ENCODE published these findings in a paper reporting on integrated analyses of ENCODE data in Nature (ENCODE Consortium 2012b) along with ~thirty companion papers in Nature and other journals (ENCODE Consortium 2012a). As part of this collection were several publishing innovations developed with Nature, including “threads,” themes that ran through multiple publications. In a remarkable demonstration of collaboration, all publications, whether they were from a journal within the Nature family or other journals (e.g., Cell, Science, J. Biol Chem.), were published simultaneously and linked to this Nature site.

With the availability of ENCODE data and tools now publicly available and an increased awareness of the project based on these Consortium publications, the utility of ENCODE as a community resource was starting to be more fully realized. In fact, it was shortly after this time that community publications using ENCODE data and tools started its exponential growth (see below and Figure 4.2).

There was some negative response to these publications, however, that largely revolved around two main points. The first was the perennial concern expressed by some in the biomedical community that “big consortium science” was a waste of resources and that the funds devoted to it would be better spent on individual investigator-directed research. This argument was not dissimilar to objections raised by the HGP and other large consortia. As discussed below, it is NHGRI’s view that there is great value in developing resources for the community that are generated in a coordinated, standardized, and efficient manner rather than being done in a distributed fashion such that comparisons across projects are not nearly as feasible.

The second main objection was the use of the term “functional” when describing the regions of genomic DNA that had been identified using the high-throughput assays employed by ENCODE researchers that primarily measured biochemical activity and, in fact, had not demonstrated definitive functional activity. This objection was fueled by the statement that the data could allow “us to assign biochemical functions for 80% of the genome, in particular outside of the well-studied protein-coding regions,” which was misinterpreted by some to mean that 80 percent has a definitive biological role. This led to a debate in the scientific community in 2013 (Doolittle 2013; Eddy 2013; Graur et al. 2013; Brunet and Doolittle 2014) about what it means to ascribe function to any genomic region.

In writing the paper, the Consortium members were aware of the sensitivities on describing ENCODE’s findings with respect to function and had tried to address it directly by stating upfront: “Operationally, we define a functional element as a discrete genome segment that encodes a defined product (for example, protein or non-coding RNA) or displays a reproducible biochemical signature (for example, protein binding, or a specific chromatin structure)” and otherwise providing qualifiers. However, our attempts were not sufficiently clear to avoid some confusion and criticism. While some negative reception, differences of scientific opinion, or objections to a scientific approach or conclusion are not unexpected, I was surprised by the intensity of the reaction by several individuals and how social media can amplify the voices of a few (which in hindsight is not as surprising).

A group of ENCODE 2 participants addressed these concerns in a paper (and follow-up letter) outlining the challenges of ascribing function to specific regions of the genome, including the lack of consensus on how to define function particularly across different scientific disciplines (Kellis et al. 2014a, 2014b). This paper discussed the strengths and limitations of the three approaches taken to defining function—experimental, sequence conservation, and genetic—describing how each provides complementary lines of evidence for function and all are needed to identify functional DNA sequences, determine their boundaries, and ascertain their biological roles at the molecular, cellular, and organismal levels.

ENCODE has made a sustained effort by to be transparent in presenting ENCODE data, analyses, resources and in communicating inferences, assumptions, and conclusions. For example, as discussed below, ENCODE compiled a “Registry of candidate cis-Regulatory Elements (cCREs)” using data from ENCODE and the RoadMap Epigenomics Project.3 In addition, a visualization tool “SCREEN” was developed to enable exploration of cCREs, how they are classified, and the supporting data (Moore et al. 2020).4

1.1.3. modENCODE and mouse ENCODE

When NHGRI Program staff presented the Concept for the ENCODE Pilot Project, the question was raised during Council discussion about why we were planning to start with the study the human genome and not first, as had been the successful strategy of the HGP, start with studying5 model organism genomes. This idea was supported by others on the Council—as noted in the Council minutes from September 2002 (National Human Genome Research Institute 2002b): “The Council approved the concept, but noted that there would also be considerable merit in a comparable program focused on model organisms.” There was also enthusiasm for such an endeavor among NHGRI staff and from that important discussion grew the modENCODE project (for model organism ENCODE).

In considering which organisms to include in modENCODE, program staff reached out to several key model organism communities for which the completed sequence was available: S. cerevisiae (yeast), C. elegans (worm), and D. melanogaster (fruit fly). Each community considered the opportunities based on the state of genomics at the time and provided NHGRI with White Papers for what such projects could look like. With community input from these White Papers, NHGRI developed two RFAs to support the modENCODE effort. These RFAs were issued in March 2006: one for data production—Identification of All Functional Elements in Selected Model Organism Genomes (National Human Genome Research Institute 2006e); and one for a Data Coordination Center (National Human Genome Research Institute 2006c) to support the generation and management of functional data from C. elegans and D. melanogaster. Based on the advanced state of genomics of S. cerevisiae and the path described for further development in the White Paper, NHGRI decided to not include this organism in modENCODE. modENCODE awards were made in April 2007, five months before ENCODE 2 was funded (National Human Genome Research Institute 2014). The modENCODE Consortium also published a White Paper outlining the goals and objectives for the project (Celniker et al. 2009).

The modENCODE Consortium members both collectively and individually published a number of papers reporting on their results. Of note are two sets of Consortium papers: The first set of papers was published online in December 2010, primarily focused on analyses within one species or the other (Gerstein et al. 2010; Ikegami et al. 2010; Ledford 2010; modENCODE Consortium et al. 2010; Allen et al. 2011; Berezikov et al. 2011; Brooks et al. 2011; Cherbas et al. 2011; Chung et al. 2011; Eaton et al. 2011; Ercan et al. 2011; Graveley et al. 2011; Hoskins et al. 2011; Kharchenko et al. 2011; Liu et al. 2011; Lu et al. 2011; modENCODE Consortium 2011; Niu et al. 2011; Nordman et al. 2011; Riddle et al. 2011; Spencer et al. 2011). The second set, published in August 2014, included both species-specific analyses as well as three major papers reporting on comparative analyses across species (Araya et al. 2014; Boyle et al. 2014; Brown et al. 2014; Gerstein et al. 2014; Ho et al. 2014). The former papers provided a detailed view about the functional components of each genome, and the latter set of papers described ways in which gene regulation in the fly, worm, and human genomes follow similar overall principles yet differ in the specific encoding and execution of their functions.

1.1.4. Impact of ARRA Funding

In 2009, NIH received significant additional funding through the American Recovery and Reinvestment Act and NIH Institutes/Centers solicited supplements and new applications (National Institutes of Health 2009).6 With this unexpected opportunity, NHGRI was able to support a Data Analysis Center focused specifically on modENCODE, which greatly enhanced the ability to analyze modENCODE data (Johnson 2011). Supplements were also made to existing modENCODE and ENCODE projects, primarily to augment data production.

In addition, three new awards were made to expand ENCODE to analyze the mouse genome, an important, well-studied model organism whose genome is more closely related to the human than fly or worm. The three groups were joined by a fourth group supported by an investigator-initiated R01 award to form the mouse ENCODE Consortium (National Human Genome Research Institute 2016e). An initial description of the resource was published in 2012 (mouseENCODE, et al. 2012). This group subsequently published 10 papers in 2014 (mouseENCODE Consortium 2014). These papers describe the data generated by mouse ENCODE as well as comparative analyses between mouse and human that identified both conserved sequence features as well as transcribed and regulatory regions that show significant divergence (Yue et al. 2014). Following mouse ENCODE, both mouse and human genomes were studied in subsequent phases of ENCODE.

1.1.5. ENCODE 3

It was anticipated that only a limited number of biological samples (mostly cell lines) would be studied by the end of ENCODE 2, and so another phase of ENCODE—ENCODE 3 (National Human Genome Research Institute 2018d)—was funded with the aim of greatly expanding the biological space to be interrogated. One RFA (National Human Genome Research Institute 2011c) supported functional element mapping in both mouse and human. A second RFA (National Human Genome Research Institute 2011b) supported a combined Data Analysis and Coordination Center with the goal that these two activities be closely coordinated. A third RFA (National Human Genome Research Institute 2011a) was a new effort to expand the computational expertise brought to bear on ENCODE data with the goal of strengthening the utility of the ENCODE resource for the research community. As in ENCODE 2, a suite of RFAs were funded in ENCODE 3 to support additional technology development (National Human Genome Research Institute 2011e–g).

During ENCODE 3, emphasis was placed on taking ENCODE data, as well as RoadMap Epigenomics data and creating a series of annotations that comprise the “Encyclopedia.”7 The Encyclopedia includes ground-level annotations of individual or related datatypes and integrative level annotations. The main product of the integrative level annotations is the registry of candidate cis-regulatory elements (cCREs) as well as a web-based tool—Search Candidate cis-Regulatory Elements by ENCODE (SCREEN)—to connect to other annotations and raw ENCODE data. The Encyclopedia is a “living” entity that will evolve and be updated, enabled by additional data (e.g., from Functional Characterization Centers [FCCs], see below) and analyses.

While the ENCODE 3 Consortium members published papers throughout the funding period, a major body of work was published in a coordinated set of papers in 2020 featuring a flagship paper and twenty-nine additional peer-reviewed papers (Nature 2020). The flagship paper provided a detailed view of the updated catalog of candidate functional elements ENCODE had compiled, including the millions of cCREs found in the human and mouse genomes, as well as using SCREEN to navigate ENCODE 3 data (Moore et al. 2020). The additional papers featured more targeted investigations into mouse epigenetics and gene expression, human gene expression and RNA regulation, and computational analysis and tools. Finally, the ENCODE 3 PIs wrote a Perspective on ENCODE to provide a brief history of ENCODE and high-level summary of ENCODE data production, genome annotation, and publications over the years (Snyder et al. 2020).

1.1.6. ENCODE 4

A fourth phase of ENCODE was funded in 2017 after receiving strong enthusiasm from the research community for continuing mapping and computational analyses as well as moving beyond cataloging candidate functional regions to functionally validating them in specific biological (National Human Genome Research Institute 2015b). As a result of these recommendations and approval of the Concept by the NACHGR (National Human Genome Research Institute 2015a), NHGRI released another series of RFAs and supported five interrelated activities: (1) continued support of mapping candidate functional elements in human and mouse (National Human Genome Research Institute 2016a); (2) new functional characterization efforts to implement a number of different approaches to test candidate elements in different biological contexts (National Human Genome Research Institute 2016b); (3) continued support of computational groups to bring additional expertise to the analysis of the ENCODE resource (National Human Genome Research Institute 2016b); (4) a Data Coordinating Center (National Human Genome Research Institute 2016d); to closely interact with (5) a Data Analysis Center (National Human Genome Research Institute 2016c). ENCODE 4 was managed similarly to previous phases of ENCODE and resulted in a greatly expanded number of mapping datasets, functional characterization of ENCODE data, and updated versions of the Encyclopedia. ENCODE 4 (National Human Genome Research Institute 2020) ran from 2017 to 2022.

Funding for ENCODE 4 ended in 2022 and another series of publications are anticipated. NHGRI developed plans to sustain the ENCODE resource for years to come, as it does not plan to fund an additional phase of ENCODE. While considerable effort is still needed to fully annotate and understand the human and mouse genomes, NHGRI’s recent strategic vision for how genomics can be applied for improving human health underscored how much genomics has evolved over recent years, warranting fresh approaches and perspectives to address these important questions (Green et al. 2020). Of note are two new efforts underway at NHGRI that address genome function in novel ways: (1) Impact of Genomic Variation on Function was funded in 2021 to systematically understand how genomic variation effects genome function and phenotype (National Human Genome Research Institute 2022b) and (2) Molecular Phenotypes of Null Alleles in Cells (MorPhiC)—Phase 1 is expected to be funded in 2022 with the long-term goal of creating a catalog of molecular and cellular phenotypes for null alleles for every human gene (National Human Genome Research Institute 2022c).

2.1.1. Interactions with Other Genomics Projects

The advantage of collaborating and synergizing with other, related, genomics projects such as the Mammalian Gene Collection (Gerhard et al. 2004) and International HapMap Project (International HapMap Consortium 2003) was also recognized to be important during the ENCODE Pilot Project. Additional collaborations were established during subsequent phases of ENCODE with several NIH Common Fund Programs, for example, RoadMap Epigenomics, Genotype-Tissue Expression (GTEx),8 and 4D Nucleome Programs,9 as well as the International Human Epigenome Consortium,10 the latter of which ENCODE was a member. The ENCODE Portal11 now hosts the RoadMap Epigenomics Program’s data where they are provided to the research community in the same processed form that is used for ENCODE data, ensuring interoperability and consistency between the Projects’ data.

2.1.2. Technology Development

NHGRI has a strong and sustained history of supporting technology development since the initial days of the HGP. When ENCODE was being planned, it was recognized that to reach the long-term goal of developing a comprehensive annotation of the human genome it would be critical to stimulate new and improved approaches, methods, and technologies to have a sufficient set of tools available for high-throughput, genome-wide identification of functional elements. To fulfill this need, NHGRI supported four rounds of funding for computational and experimental tool technology development. The first two rounds were funded during the pilot project in 2003 and 2004 and supported a total of twelve projects focused on a wide range of biochemical features of DNA including regions that were potentially functional based on their accessibility to DNA cleavage, long-range interactions between regions, and regions where regulatory proteins bind (National Human Genome Research Institute Institute 2012a). At the time of funding ENCODE 2, NHGRI supported a third round of technology development projects (National Human Genome Research Institute Institute 2018c), and the fourth was funded in 2012 at the same time as ENCODE 3 awards (National Human Genome Research Institute Institute 2018d). By the time that ENCODE 4 was being planned, NHGRI recognized that many of the technologies that were or could be developed for ENCODE had the potential for broad applicability and that methods being developed by a more general technology development call (which NHGRI was expanding) could be applied to ENCODE. Therefore, the decision was made to broaden the general call for technology development and put these efforts together under a single umbrella (National Human Genome Research Institute Institute 2018e).

In large-scale genomics projects, there is always a tension between maintaining use of existing technologies in data production throughout the course of a phase of the project versus incorporating significant technological improvements to existing methods or implementing completely new technologies that were thought to have significant benefits. Generally speaking, NHGRI was supportive of data producers implementing technical improvements to existing technologies, but it was often not feasible to switch to completely new methods during any given phase of ENCODE. However, technological advances achieved by ENCODE Technology Development grantees or by the broader research community have been incorporated into each new phase of ENCODE.

Several technologies developed with ENCODE support were sufficiently robust to be implemented in subsequent data production activities. For example, a method for whole genome bisulfite sequencing developed by Dr. Joseph Ecker’s group at the Salk Institute was used in data production of DNA methylation data in ENCODE 3. Similarly, a research team led by Dr. Jay Shendure at the University of Washington and Dr. Nadav Ahitav at the University of California, San Francisco were supported to develop a massively parallel reporter assay that was then implemented in an FCC under ENCODE 4.

Perhaps the most profound technological advance that took place near the end of the pilot project that completely changed the face of ENCODE (and well beyond) was the availability of “next generation” sequencing platforms that allowed for massively parallel sequencing, especially that provided by Illumina (see chapter 6). By 2007, shortly before the funding of ENCODE 2, sequencing machines were commercially available for researchers to implement readouts of ENCODE-style assays at an unprecedented scale and at a greatly increased specificity than the microarray-based platforms used during the pilot. Nearly all assays used in ENCODE since the pilot have employed these short-read sequencing platforms, which initially involved the successful development and adaptation of methods by ENCODE investigators. The more recently available long-read sequencing platforms from, for example, PacBio,12 were also incorporated into specific applications where the tradeoff between read length and data quality/cost was worthwhile, for example, sequencing of RNA transcripts to identify different isoforms.

CRISPR-Cas9 technologies that can introduce pre-determined and precise changes in DNA sequence made it possible to perturb sequences within candidate functional elements at high resolution and throughput (Catarino and Stark 2018). This advance supported the concept of establishing FCCs in ENCODE 4, in which a number of these centers used CRISPR-Cas9-based genome editing approaches.

2.1.3. Importance of Working on Common Targets/Samples/Reagents

The utility of having common sets of reagents such as cell lines and antibodies to facilitate comparison between different technological platforms and approaches was also identified as a key feature of Consortium interactions from the beginning, a practice that continued through ENCODE 4. In addition to serving as a means of evaluating methods and data quality, it has the tremendous advantage of enabling integration of similar as well as different data types, strengthening the ability of ENCODE to richly annotate the human and mouse genomes (see above).

2.1.4. Data Management

Overall, data management is a critically important, time-consuming and expensive component of any large genomics project, including ENCODE. Meeting the requirements of data management and analysis that the diverse data types generated by ENCODE were considered and planned for from the start of the project. The Marker Paper described the need to establish an ENCODE Portal that would serve to index the data and allow for user queries of both data and metadata. From 2003 to 2012, the primary location for community access to sequence-based ENCODE data was the UCSC Genome Browser13 with data also available at GEO.14 As the needs of ENCODE evolved, a portal and platform that could flexibly handle the volume and diversity of data was established in 2012 at the ENCODE Data Coordination Center at Stanford University, while continuing to make data available at GEO.15 This portal now hosts all ENCODE data as well as data from related projects, including modENCODE, mouse ENCODE, and RoadMap Epigenomics. It also hosts data analysis products that provide ground level as well as integrative level annotations that all contribute to the compiled encyclopedia (see above). As of December 28, 2021, 19,803 experiments were publicly available on the ENCODE portal (ENCODE Consortium 2000) with more than 14,000 generated by ENCODE. The portal also serves as a centralized source of information on ENCODE for both Consortium members and outside users, including access to publications using ENCODE data (see below), assay standards, file formats, software tools, uniform data processing pipelines, and the data use policy.

2.1.5. Data Standards

The need to develop data standards for all data types was also identified in the pilot project and continued to be a major effort as assays were improved or new assays were incorporated into ENCODE. The availability of data standards, quality metrics, experimental guidelines, and data processing pipelines ensured that high (and known) quality data were available (ENCODE Consortium 2019b). In addition, capturing detailed and well-curated metadata in a searchable form was critical to making data readily accessible to the research community. The development and use of common ontologies and controlled vocabulary both within ENCODE and shared between other related projects, such as IHEC, has enhanced reproducibility, transparency, and interoperability with other projects (ENCODE Consortium 2019c). Of the many advances in data storage and management, one key development that ENCODE has taken advantage of is the advent of cloud computing. The ENCODE DCC stores ENCODE data and metadata on the cloud through Amazon Web Services. It also makes its uniform data processing pipelines available for community use through DNAnexus (ENCODE Consortium 2019d).

2.1.6. Data Release and Accessibility

From the start, ENCODE was designated as a “community resource project” as defined in a report on genomics data sharing that came out of a meeting held in Fort Lauderdale in 2003 (Wellcome Trust 2003) (see chapter 2). The ENCODE pilot project data release policy (ENCODE Consortium 2007a) was based on the expectations laid out in that report, specifically (1) data producers will make data publicly available as soon as they are verified to be of high quality/reproducible; (2) data users should cite the source of the data by citing the Marker Paper and acknowledging the ENCODE Consortium and the data producers; and (3) data users should voluntarily afford the data producers the first opportunity to publish on the data and analyses of these data. Subsequently, ENCODE 2 and modENCODE established a joint data release policy (ENCODE Consortium 2009). The main change in this policy was to elaborate point # 3 above in more detail. It strived to be more specific about the expectations for use and publication by both data producers and data users. Data producers were encouraged to publish on their own data in a timely manner (i.e., submission of manuscript for publication within nine months of data release). Within this nine-month time frame, data users could use the data in their own analyses but were asked not to publish or publicly present on unpublished ENCODE data. After the nine-month publication moratorium (or earlier if the data had been published), data users were free to publish or otherwise report on their use of ENCODE data but were asked to acknowledge the Consortium and data producers.

In ENCODE 3 and 4, the decision was made to remove the nine-month publication moratorium, and data users are free to use data as they are released (ENCODE Consortium 2014). As before, data users were asked to reference the datasets used and acknowledge the Consortium and data producers.

2.1.7. Unrestricted Data Access

A fundamental tenet of ENCODE has been to create as broadly accessible a resource as possible. Because ENCODE serves a wide range of users, this means availability of even raw, unprocessed data in unrestricted (open access) databases such as GEO since requesting and gaining access to controlled access data in databases, for example, dbGaP, can take considerable time and effort. As ENCODE moved beyond using long-established cell lines for which a substantial amount of genomic data was already in public databases, it transitioned to using new primary cell and tissue biosamples that were explicitly consented for unrestricted data sharing (thereby allowing for general research use). The first big step in this direction came during ENCODE 3 through a collaboration with the GTEx project, named “EN-Tex,” in which approximately thirty tissue samples from each of four donors were obtained using the GTEx consent form onto which an addendum had been added that explicitly allowed for unrestricted data sharing (National Disease Research Interchange 2014). In ENCODE 4, the expectation was that all biosamples used would have been obtained using consent explicitly allowing for unrestricted data sharing. (Exceptions were allowed only in limited instances for the use of some historically used cell lines, e.g., for assays not previously used in ENCODE.) This expectation preceded the new NHGRI requirement for explicit consent for future research uses and broad data sharing that started in early 2020 (Ganguly 2019).

2.1.8. Open-Source Software

In addition to providing freely accessible data, starting with the pilot project the expectation was set forth in the RFAs for open-source software sharing. The most recent, and detailed, plan for software and analysis release was established in ENCODE 4 (ENCODE Consortium 2019d).

2.1.9. Publications and Community Outreach

Another feature of ENCODE through the years has been the publication of Consortium-wide papers as well as papers by individual participants. In keeping with the goal of serving as a community resource, the Consortium-wide papers have largely been published under Open Access agreements.

All ENCODE participants were free to publish on their own work when ready for publication, and most papers from ENCODE participants were published in this manner. However, there was always an inherent tension between meeting the needs of individual researchers to publish on their own timeline and the needs of the overall Consortium to publish as a group. While individual opinions within the Consortium have differed, the general sentiment, particularly early on, was that there was significant value in publishing Consortium-wide papers and other related papers from individuals or small groups in paper “packages,” as they served to raise the visibility of the project by notifying the broader research community about the content and utility of the resource. However, from my viewpoint, putting together Consortium-wide and associated papers came at a significant cost as well. The papers, particularly those that were Consortium-wide, always took significantly longer to write (one to two-plus years) than anticipated. Reaching consensus about topics for which there is a wide range of scientific opinion can be very challenging and time-consuming, as can coordinating analyses and companion papers to avoid unnecessary duplication and overlap. Inevitable delays often resulted in some papers being held up for long periods of time, which had the potential to negatively impact some Consortium members, especially those more junior in their careers. Further, with a project running as long as ENCODE has, with participation of some of the same researchers in multiple phases, inevitably led to some stressed relationships and publication preparation “fatigue” that have amplified these challenges. Nonetheless, publication by individuals, small groups, and the Consortium as a whole has been an important aspect of outreach to the research community as well as well-deserved recognition for work completed.

As further outreach to the community, ENCODE hosted tutorials and Users Meetings to inform the community about the available resources and how they can be used to advance the users’ own research.

2.4.1. Looking Forward

While the long-term impact of ENCODE won’t be known for years, I anticipate that it will continue to be a valuable resource for the biomedical community for the foreseeable future. Additional data, especially functional characterization, as well as new and improved analyses from ENCODE 4 will enhance the breadth and quality of genome annotations. The community will continue to use the data in novel ways to bring new insights into the genome including understanding regulation of gene expression, human health, and disease. Further, tremendous opportunities exist for understanding the influence of genetic variation on phenotype.