Inclusion Bodies

OHDSI grew out of the Observational Medical Outcomes Partnership (OMOP) [4 (link)], which was a public-private partnership established in the US to inform the appropriate use of observational healthcare databases for studying the effects of medical products. The five-year project developed new methods in observational research and established an observational research laboratory. At the conclusion of this five-year project, the OMOP research investigators initiated the OHDSI effort, and the research laboratory moved to the Reagan-Udall Foundation for the FDA under the Innovation in Medical Evidence Development and Surveillance (IMEDS) Program [5 ].
A centerpiece of the OMOP project was the development of the OMOP Common Data Model (CDM) [4 (link)] which represents healthcare data from diverse sources in a consistent and standardized way (see Table 1). This CDM is a “strong” information model, in which the encoding and relationships among concepts are explicitly and formally specified. The OHDSI team has adopted and continued maintenance of this model and its associated vocabulary services. OHDSI’s overall approach is to create an open network of observational data holders, and require that they translate their data to the OMOP CDM. Each element in the participant’s database must be mapped to the approved CDM vocabulary and placed in the data schema. In return, this approach creates a unique opportunity of implementing a number of existing data exploration and evidence generation tools and participating in world-wide studies because any given query can be executed at any site without modification. This enables multicenter, global analyses to be executed rapidly and efficiently using applications or programs developed at a single site.
Data are retained at the participant’s site, simplifying patient and business privacy issues. The team previously found that simply merging the databases is likely to give poor answers because of the heterogeneity [6 ]. Instead, the analyses are carried out locally and the results are transmitted to the coordinating center, where they can be studied on a population level and aggregated as appropriate.
OHDSI operates at several levels: infrastructure, data, methods, applications, and experiments. These levels serve both to support and inform the work of each other to ensure that the infrastructure and products support the mission. Rather than just creating a data network, OHDSI directly integrates the researchers who use the network and the data scientists who create the algorithms with the use cases for the data network.
The group’s guiding principles are that the effort be
To achieve the principle of inclusivity, OHDSI is an open collaborative. Anyone who can give time, data, or funding is welcome, and participation in the operation of OHDSI is expected. Currently, participants come from around the world, including the United States, United Kingdom, Netherlands, Sweden, Italy, Korea, Taiwan, Hong Kong, and Australia.

We loaded catalogs from over 320 commercial vendors and 130 annotated catalogs. Some sources such as HMDB and DrugBank were loaded as several distinct catalogs in ZINC allowing us to leverage the curation of metabolite origin such as plant metabolites in HMDB or drug status such as investigational drugs in DrugBank. All catalogs in ZINC are categorized by their biogenic and bioactivity status, if any.⁹⁵ Only descriptions that characterized the entire catalog contents were applied. For instance, the “Approved” subset of DrugBank was categorized as “World Drugs” since it contains over 100 drugs approved in other countries but not by FDA, and the “Endogenous” subset of HMDB was categorized as having a biogenic type of “endogenous human metabolite”. Molecules inherit biogenic and bioactive properties from the catalogs they are found in. These values are computed and stored, and are accessible in the interface as molecular features. There are four biogenic catalog levels: 1) Endogenous human metabolites, i.e. compounds that are synthesized in man. Interestingly, this may include compounds produced by our bacterial flora; 2) Metabolites of any species, i.e. small molecules that are involved in metabolism, development and reproduction, but not metabolites of xenobiotics; 3) Biogenic compounds, often called natural products; 4) Unknown biogenic status. Likewise, ZINC supports seven levels of bioactivity annotation as follows. 1) FDA approved; 2) World drugs; 3) Investigational, compounds reported to be used in clinical trials; 4) In Man, which including nutraceuticals, for instance; 5) In vivo, which includes DrugBank experimental compounds that have been in animals; 6) In cells, which includes compounds reported active in cell based assays; 7) In vitro, compounds active or assumed active at 10 μM or better in a direct binding assay. All other compounds are marked as having unknown biological activity. The categories are ordered to be progressively inclusive within each series, thus all FDA approved drugs are also world drugs and all compounds active in cells are also active in vitro. We annotate as building blocks those catalogs of compounds available in preparative quantities, typically 250 mg or more. Commercial vendors are categorized by the speed and cost of compound acquisition, allowing the best purchasability of every compound to be computed based on its current catalog membership. Catalog categorizations are refined continually by purchasing experience in our lab and reports from colleagues, as follows:⁹⁵ 1) In stock, delivery in under two weeks, 95% typical acquisition success rate; 2) Procurement agent, in stock, delivery in 2 weeks, 95% typical acquisition success rate; 3) Make-on-demand, delivery typically within 8 to 10 weeks, 70% typical acquisition success rate; 4) Boutique, where the cost may be high, but still likely cheaper than making it yourself, 70% typical acquisition success rate.