An EPA database on the effects of engineered nanomaterials-NaKnowBase

The US EPA Office of Research and Development (ORD) has conducted a research program assessing potential risks of emerging materials and technologies, including engineered nanomaterials (ENM). As a component of that program, a nanomaterial knowledge base, termed “NaKnowBase”, was developed containing the results of published ORD research relevant to the potential environmental and biological actions of ENM. The experimental data address issues such as ENM release into the environment; fate, transport and transformations in environmental media; exposure to ecological species or humans; and the potential for effects on those species. The database captures information on the physicochemical properties of ENM tested, assays performed and their parameters, and the results obtained. NaKnowBase (NKB) is a relational SQL database, and may be queried either with SQL code or through a user-friendly web interface. Filtered results may be output in spreadsheet format for subsequent user-defined analyses. Potential uses of the data might include input to quantitative structure-activity relationships (QSAR), meta-analyses, or other investigative approaches.


Background & Summary
The recent advances of nanotechnology have led to concerns for the potential release of engineered nanomaterials (ENM) into the environment causing exposure to, and perhaps adverse effects on, humans or sensitive ecological species 1 . Accordingly, the United States Environmental Protection Agency (US EPA) Office of Research and Development (ORD) has developed a research program aimed at understanding the potential environmental implications of ENM. ORD research encompasses potential releases of ENM from manufacturing and commercial uses; environmental transformations, fate, and transport; exposures; and potential adverse health effects. A framework was developed to organize and integrate this diverse set of information 2 . To support this larger effort, a relational database was developed containing ORD nanomaterial research data to better enable the use and synthesis of study results, and to facilitate higher-order analyses such as quantitative structure-activity relationships (QSAR). One goal is to probe the relationships between physical and chemical properties of ENM and their environmental actions to see if predictive relationships can be determined. This publication announces the release of "NaKnowBase" (NKB), a knowledge base containing the results of multiple ORD publications on the actions of ENM in environmental or biological media.
The design of NKB was intended to compliment efforts in nanoinformatics -the strategic curation and collation of nanomaterial data for analytic purposes. A roadmap for nanoinformatics in the European Union (EU) and US was recently published providing a comprehensive overview of the inter-related scientific disciplines of nanomaterials science, physicochemical characterization, computational modelling, informatics, and ecological and human toxicology 3 . This analysis identified three challenges facing nanoinformatics: (1) limited datasets, (2) limited data access, and (3) regulatory requirements for validating and accepting computational models.
NKB partially addresses the first two of these issues by providing a publicly available source of curated data relevant to ENM environmental health and safety (EHS). Collating datasets from multiple sources facilitates more comprehensive meta-analyses, QSAR, and risk assessment approaches such as read-across 4 . To date, such "big data" endeavours in ENM EHS tend to be designed around large datasets that must be generated in advance, or remain limited by a paucity of relevant, curated data from disparate sources [4][5][6] . Efforts like NKB can help overcome these research hurdles by being strategically designed to leverage extant data while also being amenable to newly generated data.
There are other nanomaterial-related databases indexed in the appendix section of the EU-US roadmap 3 . These databases are independently operated and vary according to the intended use and operability, the types of data captured, and the data format, access, and control. Although it may appear advantageous to consolidate these, there are several factors favouring the maintenance of independent databases: ability to control access to, quality of, and integrity of the data, managing and protecting proprietary and confidential business information, the pragmatics of scale, and the availability and continuity of funding. Therefore, the original scope of the NKB was limited to data collected by the EPA ORD. To our knowledge, the data provided in NKB are not collected elsewhere. The data in NKB represent the only collated source of published data from the US Environmental Protection Agency in a relational database regarding the potential environmental effects of engineered nanomaterials.
NKB was built as an SQL relational database. The overall structure is shown in Fig. 1. The database has separate tables on the source publication, the tested materials and their physicochemical properties, the media in which the materials were tested, the assays performed, the parameters evaluated, and the results. There are sub-tables to capture data on chemical contaminants, attached functional groups, and test media additives. Data entry is accomplished by curators via a set of prescribed Excel spreadsheets that are then imported to the database using a script. During curation, efforts are maintained to use terminology consistent with an expanded nanomaterial ontology being developed by several nanoinformatics groups including the EU NanoSafety Cluster and the Center for the Environmental Implications of Nanotechnology (CEINT), in coordination with the foundational work published by the eNanoMapper database 7,8 . In addition, a simple, user-friendly interface was developed which allows users to search the database and obtain outputs of data in spreadsheet format.

Methods
Publications selected for curation were limited to research conducted by ORD and related to environmental or biological actions of ENM. This included in vivo, in vitro, and in silico experiments as well as life-cycle analyses and physicochemical characterisations. The data in the database reflect over 120 relevant publications from approximately 2012 through November 2019. Over 70 unique nanomaterials as defined by the combined composition of the core, shell and coatings were studied. Over 160 named assays and 22,000 individual assays were run. We expect to maintain the database and continue to make additions over time as new research becomes available. Though NKB will be made available through the Office of Science Management as a public EPA database tool, pertinent NKB data will also be integrated with the CompTox Chemicals Dashboard (https://comptox-prod. epa.gov/dashboard/chemical_lists/), which maps the DSSTox substance records to the most current list of NKB nanomaterials. The addition of new data will be announced via the CompTox Chemical Dashboard (https://comptox.epa.gov/dashboard/) on the 'News' (https://comptox.epa.gov/dashboard/news_info) and 'Downloads' (https://comptox.epa.gov/dashboard/downloads) pages of the Dashboard, as appropriate.
The EPA maintains various repositories for planned, ongoing, and completed research and projects. These repositories were searched for relevant publications for curation. The description and content of these repositories are detailed below. Fig. 1 Overview of the NKB SQL structure. The lines indicate the nature of each relationship. Each relationship is of a one-to-many nature, where the end with two lines is "one" and the end with a triangle is "many", such as one publication being able to have many mediums. Direct input from investigators. Where available, ORD researchers provided their publication(s) and original data for inclusion in the database. These papers and submitted data were evaluated on a case-by-case basis and formatted by trained curators for inclusion in the database. Approximately 9% of the entries were submitted directly by the investigators. Among the reasons that original data may not have been available included the primary investigators having left the Agency, data having been archived, lack of access to raw data from scientific instruments, and incompatible formats. An example of an incompatible format was lists of differentially expressed genes encoded as "increased" or "decreased" where the data fields in the NKB required numeric value entries.

Systematic article selection.
Papers of interest were identified by running keyword searches through STICS, Science Inventory and Science Hub. A list of entries containing "nano" in the keywords or title were obtained. Additional queries were run separately using search terms including the composition of common ENM (e.g. silver, copper, titanium dioxide, cerium dioxide, etc.). Results were checked for duplicates, and posters, abstracts, or meeting presentations were not considered for curation. Over 600 titles were identified for further screening. These results were then reviewed to identify only original, peer-reviewed research. Finally, titles and abstracts were carefully read for relevance to nanotoxicology, environmental effects of nanomaterials, physical and chemical properties, and ENM life cycle. Other nanomaterial papers including literature reviews and those relating to topics such as incidental or naturally occurring nanomaterials, method development or "green chemistry" synthesis of nanomaterials were excluded.

Table organization and curation procedures.
The curation of data into the database required a set of trained data curators and a substantial commitment of time and effort. Artificial intelligence or other automated procedures were not used. The original training of data curators was generously conducted by the database experts of the Center for Environmental Implications of Nanotechnology (CEINT) in association with the Nano Informations Common (CEINT NIC), a database maintained at Duke University in Durham NC. Experienced NKB curators subsequently oversaw the training of new data curators as needed. Training consisted of explaining the overall purpose and structure of the database and the data input templates, and then overseeing the curation of selected model datasets which had been curated previously by others. When the novice curators were sufficiently proficient at capturing data from the training sets, they began with oversight to encode new manuscripts. Curators typically became proficient in a matter of a few weeks. Once curators were proficient, curation of data from each new manuscript typically required between one to several workdays depending on the complexity of the material. Questions or uncertainty about experimental procedures or parameters were referred to the project management and occasionally required contact with the authors of the original manuscripts for clarification. Thus, the robust curation of data for the database required considerable time and effort of skilled personnel.
Data extraction and curation occurred in accordance with an approved EPA quality assurance project plan (QAPP E-TAB-0030177, Project ID "Emerging Materials Project 18.02"). In summary, all data were collected from published journal articles. Metadata were attached to all curated data. Data were extracted from  Table 3. Medium table data fields. Data regarding the dispersion mediums and any additives to the mediums were recorded. The Medium and Additive tables were used to track information on any medium a nanomaterial was suspended in during an experiment. Where multiple instances of media changes or particle characterizations were made over time, these were recorded in association with experimental time variables with appropriately linked experimental parameters. Mediums were uniquely identified by a combination of their source publication's DOI and an incrementing number, MediumID, since one research publication could have studied multiple mediums. Complete medium data included the unique identification key and a description of the medium, such as a common name or a majority component (e.g. Dulbecco's Modified Eagle Medium or water).

Description AdditiveID
Unique (within publication) numerical identifier to link an additive to entries in other tables   . Shape recorded the typical shape of the material, which was important for materials like carbon that varied wildly (e.g. sheets, tubes, or a simple bulk form). If the material was suspended in a medium, that medium was referenced by DOI and Medium. This allowed for important rows about medium-specific qualities, such as Shape in Medium or Solubility, to be captured. Specifically, NKB captures many quantitative characteristics for a nanomaterial, e.g. outer diameter, inner diameter, length, thickness, surface area, size distribution, purity, hydrodynamic diameter, and surface charge. Many publications report these data using summary statistics without raw data. Therefore, each ENM characteristic was described using a set of seven fields capable of capturing raw and processed data: Value ApproxSymbol, Unit, Uncertainty, Low, High, and Method. Average contained either the raw or average numeric value reported for a measurement. ApproxSymbol captured any qualifying characters (e.g. <, >, ~) denoting measurements that lacked precision, typically due to a limitation of the machine used for measurement. Unit contained the physical unit for the measurement, using standard scientific abbreviations when possible. Raw data were reported using these first three fields along with Method. The Uncertainty, Low, and High fields are used in combination to describe the spread or distribution of processed data. The Uncertainty field held statistical terms such as "range" or "standard deviation". If the term required two endpoints, Low and High held the numeric values for those respective endpoints. For example, the Low and High of an "interquartile range" would be the first and third quartile values, respectively. If the "Uncertainty" statistic term required only one value (e.g. standard deviation), the value was recorded in Low. Finally, the technique or method used to produce the raw or processed measurements was recorded in Method (e.g., transmission electron microscopy).

Concentration of additive
www.nature.com/scientificdata www.nature.com/scientificdata/ the primary key; the Material, Assay, and Medium tables use two keys. Primary keys and foreign keys are used to connect related data that are stored in different tables.
NKB User interface. The NKB user interface application is currently under development. Deployment is expected in 2023 under the EPA web domain naknowbase.epa.gov. Here, curated data can be accessed through a user-friendly interface and search results can be downloaded for subsequent analysis by the user. NKB data can be filtered by numerous parameters such as ENM composition, physical and chemical characteristics, assay name and type, assay parameters, and result name. NKB data points are also linked to the original peer-reviewed publications via a single hyperlink.
The NKB user interface allows users to search for data using a pre-defined list of relevant search terms categorized by data tables and table fields. The searchable data fields were derived from those listed in Tables 2-11.

ContamID
Unique identifier for the contaminant data point.  Table 6. Contam table data fields. The contaminants table, "contam", served as an addendum to the material table. The primary key was comprised of the publication DOI and MaterialID of the contaminated material. The field Contaminant listed the name of the contaminating substance. ContamAmount, ContamUnit, and ContamMethod held the information on the scale of the contaminant and the way the contamination was measured. This allowed for a material to have any number of contaminants, each detailed in its own row.

MaterialFGID
Unique identifier for the functional group-material link.
material_MaterialID Material ID of the material which has a functional group attached material_publication_DOI Reference to DOI of source publication functionalgroup_FunctionalGroup Chemical identity of the functional group.

FunctionalizationProtocol
Technical method to functionalize the material (e.g. acid wash, etc.) Table 7. Materialfg table data fields. Materialfg connects specific functional group data to the broader material data. If a material had functional groups, these were tracked in the functional group and materialfg tables. Functional group was a simple list of predefined functional groups. Each row in the materialfg table was a combination of a functional group, a material ID, a publication DOI, and the name of the functionalization protocol used to add the functional group to the material. A material could have any number of functional groups.   Figure 1 and Table 1 describe all the individual data sources integrated in NKB. The NKB data frame has been uploaded into a single collection entitled "NaKnowBase-SQL backend-080121" 9 . The files contained in this collection include the most recent SQL data structure for NKB, including all tables, as well as corresponding data categories and keys for the backend of the database.

Data Records
EPA nanomaterials present in NKB are also provided through the CompTox Chemicals Dashboard (https:// comptox.epa.gov/dashboard/chemical-lists/NAKNOWBASE), which maps EPA chemical substance records to the most current list of NKB nanomaterial substance records (last updated 12/14/2020).

Technical Validation
In general, there are many varied methodologies for cataloguing nanomaterials metadata and physicochemical properties; NKB attempts to capture as much of this information as possible.
Publications considered for curation were limited to ORD research, which is subject to rigorous internal and external quality control and peer review. All research conducted at ORD must have a corresponding Quality Assurance Project Plan (QAPP). QAPPs describe the necessary quality assurance and quality control measures needed to produce results that meet stated performance criteria. ORD OAPPs are peer-reviewed, approved by management, overseen by a quality assurance manager, and subject to periodic QA and performance quality checks. Manuscripts submitted for publication are linked to approved QA plans and are subject to QA review and approval. Furthermore, manuscripts are subject to thorough internal scientific peer review before undergoing additional external, independent peer review by the publishing journal. These systems are intended to ensure the quality and accuracy of ORD data, and help assure the reliability of data being curated in NKB. Because of this, the results of the papers themselves were not checked for errors during data curation. Instead, quality control efforts focused on ensuring the accuracy of the curated data compared to the original raw data, as well as consistent curation procedure between curators.   www.nature.com/scientificdata www.nature.com/scientificdata/ To assess the quality of NKB curation, a random sampling (approx. 5%) of curated papers were manually checked for quality control. It was found that data derived from the digitization of published graphs differed from the original data by an average of 0.20% ± 0.29% (N = 316) and that curation of the same data by different curators differed by an average of 0.33% ± 3.3% (N = 736). The data are calculated as Mean ± SD normalized to the axis scale.

Usage Notes
Potential uses of the data include input to quantitative structure-activity relationships (QSAR), meta-analyses, or other modeling or investigative approaches. Users should be aware that data obtained from the NKB includes a large number of potential parameters related to physicochemical properties of ENM. Because relatively few of these properties were entirely consistent across sources, the NKB contains many sparsely populated fields. Users should consider this when planning analyses of data from the NKB. Updates to the NKB described herein help inform new testable hypotheses about the etiology and mechanisms underlying ENM effects in the environment and adverse health outcomes of toxicological concern in relation to human exposure to nanomaterials.

code availability
All custom code created to process of manipulate external datasets in the construction or subsequent update of the NaKnowBase relational database tables are made publicly available by the U.S. Environmental Protection Agency, Office of Research and Development (ORD) 9,10 . Table 11. Molecularresults table data fields. This table was an alternative to Results used to store references to results that exceeded the capacity of NKB for complexity such as genomic, proteomic or metabolomic assays. Such results were typically already deposited on outside data repositories. NKB, in these cases, provided the ENM specific aspects and experimental design considerations of these studies, which could be linked to the large datasets housed elsewhere. Rows in this table catalogued web addresses to external sources for the results in question.