BioChemUDM: a unified data model for compounds and assays

Compound identification

A fundamental part of a pharmaceutical information platform, in general terms but more specific to this application is the compound registry. The registry provides a unique identifier for a compound, which enables other entities (batch, sample) as a test article in an assay experiment. Establishing a registry involves business rules for drawing sketches, standardizing compounds, normalizing forms, and handling enhanced stereochemistry [17]. As proposed by Hersey and others [18], we subscribe to use a process consistently rather than attempt to curate chemical databases across collaborators. With this in mind, we employed a process-driven registration approach that enables data exchange between organizations where identical compounds are drawn differently.

The BioChemUDM is based purely on the chemical connection table with enhanced stereochemistry, so a compound identifier is based solely on a sketch . With one assumption about stereocenters (unspecified means mixture), there is no need for complicated business rules or controlled vocabularies to distinguish a compound. Interpretation of sketches using normalization techniques lends itself well to the language of the chemist, who is skilled in drawing molecules. Upon compound registration of a sketch, the following fields are key for data sharing.

The field ‘Compound’ is part of the molecule (MOL) object:

• UDM.MOL.Compound – Connection table with enhanced stereochemistry.

The field ‘Num’ is the unique identifier for the compound:

• UDM.MOL.Num – Registry number assigned by the system, e.g., 7.

The field ‘Name’ is a globally unique identifier, typically denoted by an organizational prefix:

• UDM.MOL.Name – A organizational prefix with registry number, e.g., IDYA-7.

To enable cross-references between multiple organizations, the field ‘Synonyms’ extends the MOL object of the BioChemUDM:

• UDM.MOL.Synonyms – An array UDM.MOL.Name values (from other registries).

To allow for verification of the compound, the MOL object represents additional unique identifiers:

• UDM.MOL.InChI – A non-proprietary identifier for chemical substances.

• UDM.MOL.InChIKey – A compact chemical identifier derived from InChI.

• UDM.MOL.IUPAC – A name for organic compounds as recommended by IUPAC.

One aspect of compound identification that deserves attention is handling of tautomers. To wit, a study of commercially available samples shows the same compound may be sold as different products at different costs [21]. Tautomerism is complicated [19] and steps to address tautomers in compound registration have been described previously [20]. Our normalization technique for tautomeric states entering the UDM is based on recent work from Dhaked and others [22]. For each compound, a set of business rules, i.e., SMIRKS patterns [23], is applied to afford a canonical tautomer for registration. These patterns, originally developed for use with the CACTVS tools [24], are adapted for use in RDKit toolkit [25] and Pipeline Pilot™ (PLP) software [26]. Furthermore, recent work by Baker and others demonstrates that use of an energy function to establish these business rules may yield tautomers in better agreement with experimental observations [27]. See Appendix B for details on the canonical tautomer implementation.

Drawing rules

There are three kinds of enhanced stereochemistry sketches: mixture, relative, and absolute.

The sketches shown in Fig. 1 naturally support the kinds of compounds synthesized during compound progression. We may begin with a stereochemical mixture, then progress to a separation of isomers, and end with a determination of chirality. In practice, these three kinds may be implied by the following sketches in Fig. 2.

Fig. 1:

Sketch of mixture (DL), relative (D or L), absolute (L-alanine).

Fig. 2:

Sketch of straight bond, wiggly bond, wedge/hash bond.

Using an industry standard drawing tool [28], some chemists naturally draw standard sketches (Fig. 1) while others prefer shorthand sketches (Fig. 2). Therefore,

1. a “straight bond” means a mixed stereocenter

2. a “wiggly bond” means a stereocenter is enantiopure but not determined

3. a chiral center without designation means it is absolute.

Compound KeyVal

Using the Morgan Algorithm [29], CIP rules [30], and PLP scripting language [31], we produced a ‘Perceive Structure & Stereo’ component to promote straight bonds, enumerate wiggly bonds, and ensure proper representation of relative stereochemistry, pseudo-stereoisomers, atropisomers, global chirality, and meso compounds. The PP component prevents invalid registrations and is amenable to implementation in KNIME™ script [32]. The component exposes a categorical chirality attribute called StereoKeyVal for informational purposes.

The field ‘StereoKeyVal’ is part of the MOL object:

• UDM.MOL.StereoKeyVal – Concatenation of chirality attributes (output of perception).

  • Format: {chiral} {isomer} {ee} {member} {meso}

  • {chiral} – T or F, depending on whether the compound is chiral (rotates light).

  • {isomer} – A, D, E, N

    • A – Atropisomer

    • D – Diastereomer

    • E – Enantiomer

    • N – Not a stereoisomer

  • {ee} – 0 or 1, a Boolean depending on whether stereo excess can define a mixture.

  • {member} – M or P for atropisomers, A or B for pseudo-stereo compounds, otherwise X.

  • {meso} – T or F, depending on whether the compound is a meso form (no mirror image).

Batch identification

For a given compound, many batches may be synthesized. So, the BioChemUDM for batches depends on the compound and each batch is identified by registration order. For example, the seventh compound registered would be IDYA-7 and the third batch of that compound would be IDYA-7-3.

The field ‘Num’ on the Batch (BAT) object is the unique identifier for the batch (in this registry):

• UDM.MOL.BAT.Num – Sequential number assigned by the system, e.g., 3.

The field ‘Name’ is a globally unique batch identifier, by extending the compound identifier.

• UDM.MOL.BAT.Name – The compound name and the batch integer, e.g., IDYA-7-3.

The field ‘ExperimentID’ is a reference to an electronic laboratory notebook (ELN) experiment.

• UDM.MOL.BAT.ExperimentID – The link to the source ELN experiment.

The field ‘Source’ is the owner/author of the ELN experiment.

• UDM.MOL.BAT.Source – The controlled vocabulary (list) of collaborators.

The field ‘Tag’ is the organizational project code to associate a batch with a project.

• UDM.MOL.BAT.Project – The controlled vocabulary (list) of projects.

To ensure an immutable batch identifier, we do not include salt form as part of the batch identifier thus avoiding the inevitable problem of an incorrect salt form assignment. Instead, we opt to define salt form as an independent data label (UDM.MOL.BAT.Salt). See Appendix C for the complete list of batch fields.

Sample identification

For a given batch, many samples may be containerized, thus, the BioChemUDM for samples depends on the batch and each sample is identified by registration order. In this way, an immutable sequential number starting at 1 can be used for compounds, batches, and samples and everything else is an attribute thereof.

The field ‘Num’ on the Sample (SAM) object is the unique identifier for the batch (in this registry):

• UDM.MOL.BAT.SAM.Num – Sequential number assigned by the system, e.g., 1.

The field ‘Name’ is an extension of the globally unique compound identifier:

• UDM.MOL.BAT.SAM.Name – The batch name and sample integer, e.g., IDYA-7-3-1.

The field ‘ID’ is the barcode associated with the sample container.

• UDM.MOL.BAT.SAM.ID – The barcode associated with the sample container.

The field ‘Type’ is a controlled vocabulary (list) of container types (e.g., dram1, dram4).

• UDM.MOL.BAT.SAM.Type – The type of sample container

The field ‘Location’ is a controlled vocabulary (list) of inventory endpoints.

• UDM.MOL.BAT.SAM.Location – The current location of the container.

The field ‘Form’ is a controlled vocabulary (list) describing the formulation of the sample (e.g., neat, solution).

• UDM.MOL.BAT.SAM.Form – The formulation of the sample.

The field ‘Amount’ is the quantity of a neat sample (without weight of the container).

• UDM.MOL.BAT.SAM.Amount (mg) – The sample quantity measured in milligrams.

The field ‘Volume’ is the quantity of a solution sample.

• UDM.MOL.BAT.SAM.Volume (mL) – The sample quantity measured in milliliters, used when UDM.MOL.BAT.SAM.Form=solution.

The field ‘Conc’ is the concentration of a solution sample.

• UDM.MOL.BAT.SAM.Conc (uM) – The sample concentration measured in micromoles/liter, used when UDM.MOL.BAT.SAM.Form=solution.

Assay identification

Another fundamental part of the information platform is the assay definition. Ontology concepts serve as intentionally broad categories for organizing assays. We propose the following protocol categories for organizing pharmaceutical data:

1. Activation

2. Binding

3. Induction

4. Inhibition

5. Oxidation

6. Permeability

7. Pharmacokinetics

8. PhysChemProperty

9. Stability

10. Toxicity

Protocol categories are beneficial because similar assays can be treated in the same way. Within each category are fields to distinguish one assay from another. For example, two kinds of activation are distinguished by the conditional field called ‘Target’ (set to AhR or PXR). To simplify the problem, we represent unpivoted data within the categories.

Each protocol category includes a common set of fields:

• UDM.ASY.Name – The controlled vocabulary (list) of protocol names (shown above).

• UDM.ASY.Protocol – The author and version number of the protocol, e.g., IDYA/v1.

• UDM.ASY.Date – The run date of the experiment. For multi-day experiments, the start day.

• UDM.DOC.Query – A query string reference to documents associated with the experiment.

• UDM.ELN.Query – A query string reference to the source ELN experiment.

• UDM.COM.Contact – A controlled vocabulary (list) of internal points of contacts.

• UDM.PRJ.Code – A controlled vocabulary of project identifiers.

• UDM.ASY.Identifier – A controlled vocabulary (list) of assay identifiers.

• UDM.ASY.TST.Article – A test article, e.g., a compound, batch, or sample identifier.

• UDM.ASY.DAT.Control – A reference to a control test article.

• UDM.ASY.DAT.Value – The readout for the control.

• UDM.ASY.DAT.Comment – A description of the control value or unit of measure.

• UDM.ASY.CND.Format – A controlled vocabulary (list) of assay formats.

• UDM.ASY.CND.Target – A controlled vocabulary (list) of assay targets.

Each protocol contains result types and specific conditions. For example, the remainder of the ‘Activation’ protocol is:

• UDM.ASY.CND.Conc (uM) – The concentration of the additive/stimulant (specified in the UDM.ASY.CND.Format field)

• UDM.ASY.RES.Emax (%) – The maximum fold activation.

• UDM.ASY.RES.FoldActive – The fold-change over baseline.

• UDM.ASY.RES.EC50 (uM) – The concentration associated with half-maximal effectiveness.

The ‘Activation’ protocol defines two assays: AhR and PXR. See Appendix D for the complete list of protocol fields, controlled vocabularies, and templates.