Introduction: Consequent to a memorandum of understanding between the Karolinska Institutet and the International Union of Basic and Clinical Pharmacology (IUPHAR) in 2018 a report on academic drug development, including guidelines (ADEV) has been drafted [1]. As part of this exercise, we conceived a triage for comprehensive informatics profiling around the compound, target, disease axis. We have termed this “in slico 360” (INS360) the aim of which was to support ADEV teams since they may lack either internal expertise or external support to do this on their own. Indeed, some past SciLifeLab Drug Discovery and Development Platform projects had been halted because of overlooked competitive impingements or insufficient target validation evidence. Methods We assessed the current database landscape, mostly public but including commercial, for potential utility for INS360. We were guided primarily by content coverage, usability, and reputation. We also explored some open property prediction resources for assay interference and toxicological inferences. Results: As a first-stop-shop, we selected the IUPHAR/BPS Guide to PHARMACOLOGY with ~900 ligand-target relationships captured via expert curation of journal papers Moving up in scale we evaluated ChEMBL at 1.8 million compounds with 1.1 million assay descriptions and 7,000 targets. With yet another jump we could search the patent corpus with 18 million extracted compounds in SureChEMBL. We explored PubChem that integrates these three with over 500 other sources linked to 96 million compounds, BioAssay results and connectivity into the NCBI Entrez system. The final jump in scale for document-to-chemistry navigation was represented by SciFinder with 155 million structures. On the target side, 360-exploration has the need to encompass literature, structure, genetic variation, splicing, interactions, and disease pathways. From their UniProt links, both GtoPdb and ChEMBL provide these entry points. Navigating genetic association data in support of target validation was enabled by the OpenTargets portal and the GWAS Catalog. We also fount servers that could produce prediction scores from chemical structures for a range of features important for de-risking development. Conclusion: This work scoped out initial resource choices for the INS360. We propose that not only ADEV operations but essentially any pharmacology research team has much to gain from this approach and many potential pitfalls can consequently be avoided when approaching key checkpoints, such as preparing a publication. However, support may be needed for both institutions and teams to get the best out of these complex and feature-rich databases. [1] Southan C, (2019) Towards Academic Drug Development Guidelines, ChemRxiv pre-print no. 8869574

1. An in silico 360 concept for the triage and
selection of drug development compounds
Introduction
As commissioned by the International Union of Basic and Clinical
Pharmacology (IUPHAR) and SciLifeLab/KI a 2018 report on
academic drug development (ADEV) is available as a pre-print
(https://europepmc.org/article/PPR/PPR102764)
This includes a triage conceived for comprehensive informatics
profiling around the compound – target - disease axis. We have
termed this “in silico 360” (INS360) to enable ADEV teams.
Notably, some past SciLifeLab Drug Discovery and Development
Platform projects have been halted because of overlooked
competitive overcrowding. INS360 can give risk indications such
as; lack of genetic validation, chemical liabilities, target promiscuity
or safety signals. These warnings are crucial for ADEV and
pharmacology research in general. As a scoping exercise, we
assessed the increasingly crowded database landscape for resources
potential INS360 utility. This was guided by judgment (necessarily
with a subjective element) on reputation, sustainability, content and
usability.
Overview Scheme
The conceptual divisions we produced for INS360 are below.
Given such a broad scope, we can only cover a selection of
resources here. As implicit recommendations these are made in
good faith but there many useful alternatives. Indeed, determining
the comparative utility between databases is a major challenge of
the informatics ecosystem.
Exploring Chemistry
Searching against bioactive structures, inactive control compounds,
drugs, leads and probes is a pivotal step for INS360. We narrowed
this down to the four sources below with their (Nov 2019)
compound numbers.
We recommend the IUPHAR/BPS Guide to Pharmacology
(GtoPdb) as a first-stop-shop for both ligand and target searches
since it is small (because of curatorial selectivity) and is designed
for pharmacologists by pharmacologists (see PMID: 31691834). At
just over 200x the size, ChEMBL extends coverage but this comes
with increased navigational complexity. This rule of thumb also
holds for the next two jumps in scale.
Christopher Southan, Biomedical Sciences, University of Edinburgh and TW2Informatics, Göteborg,
Sweden.
Per I Arvidsson, Science for Life Laboratory (SciLifeLab) and Karolinska Institute (KI), Stockholm,
Sweden.
The 710 sources in Pub (including GtoPdb and ChEMBL) are
linked to 268 million BioAssay results and connectivity to PubMed
via the NCBI Entrez system. The final quantitative leap for
document-to-chemistry linkage is represented by SciFinder at ~ 160
million (a commercial product but widely available in university
chemistry departments). Both PubChem and SciFinder contain
chemistry from patents the searching of which is an essential (but
often neglected) part of INS360 (see PMID: 26194581).
Exploring Protein Targets
INS360 needs to deeply interrogate protein sequence and structure
space for not only human entries but model organisms from yeast to
Zebrafish to rodents. UniProt and its manually curated subset
Swiss-Prot and PDB directly are first-choices here because they
include expert annotations, and ~ 100 cross references to literature,
structure, splicing, protein interactions, and disease pathways (both
GtoPdb and ChEMBL link reciprocally to UniProt). The statistics,
shown below, indicate the impressive coverage, including
increasing numbers of GPCR structures crystalised with bound
ligands.
Genomic and Literature Portals
For navigating the genome, including the comparative genomics of
target evolution and genetic disease association data, we
recommend Ensembl as the entry point. Exploration can them
extend to the two complementary and information-rich druggable
genome portals of Open Targets and Pharos (via Ensembl or
UniProt out-links)
Last but not least, we can consolidate INS360 by direct literature
searching (one could also start with this since the overview schema
can be entered from different points). Our choice of European
PubMed Central (EPMC) is a considered one in view of recent
updates increasing integration with other European resources.
Conclusions
Pharmacology will see an increasing impact from INS360 as AI
becomes an adjunct to data mining. It should be implemented (and
refreshed) at key project checkpoints such as; a new target,
optimising chemistry, moving into in vivo testing, preparing
publications, funding applications or seeking out-licencing.
However, expertise spanning different domains is needed for
cogent interpretation and exploitation. Crucially, this should also
extend to experimental confirmation of key in silico findings that
(by definition) are predictions or inferences but not “ground truth”.
Because execution of INS360 may lie outside the competences of
even experienced experimental pharmacology teams external
support will be needed. However, unlike pharmaceutical companies
who have enterprise solutions, academics may not have equivalent
facilities. Addressing this challenge is beyond this poster but we can
highlight the ELIXIR organisation. This integrates life science data
and training resources from across the UK and Europe, including
some of those mentioned above (https://elixiruknode.org/)
Flash poster 17 Dec 11.00