Platform release graphic showing the releases as a series of connected cogs

Open Targets Platform 21.09 has been released!

Release Notes Sep 30, 2021

The latest release of the Open Targets Platform — 21.09 — is now available at

Key points

Key stats

Metric Count
Targets 60,636 (of which 29,531 are associated with at least one disease)
Diseases 18,663 (of which 15,434 are associated with at least one target)
Drugs 12,594
Evidence strings 11,071,233
Associations 7,927,820

The main focus of this release has been to improve target annotation, and so we have made a number of changes to the target profile pages, including new tractability data, a new widget for genetic constraint data, and improvements to our literature mining strategy. We also have some exciting data updates.


The PROTACtable Genome: assessing drug target tractability
An Open Targets project has created a new framework to assess whether human proteins could be targeting using Proteolysis Targeting Chimeras (PROTACs).

Following the release of a new framework to assess whether human proteins are amenable to targeting using Proteolysis Targeting Chimeras (PROTACs), the tractability widget now features an assessment of the target’s ‘PROTACtability’. The PROTACtability assessment is available for 19,492 targets.

We have also redesigned the tractability widget to make the available information cleaner and more intuitive.

Widget featuring the title "Tractability", and the columns "Small molecule", "Antibody", "PROTAC", and "Other modalities"
The tractability widget on the GRIN3A profile page. The new design gives a straightforward overview of the tractability options for this target.

Genetic constraint data

The Platform features a new datasourcegnomAD — from which we ingest genetic constraint data. Genetic constraint is a way of measuring the importance of a gene to an organism’s function, by comparing the number of inactivating mutations observed in natural populations to the number of expected (predicted) mutations. Genes that are essential to human function will tolerate fewer inactivating mutations than non-essential ones.

The mutational constraint spectrum quantified from variation in 141,456 humans - Nature
A catalogue of predicted loss-of-function variants in 125,748 whole-exome and 15,708 whole-genome sequencing datasets from the Genome Aggregation Database (gnomAD) reveals the spectrum of mutational constraints that affect these human protein-coding genes.

The Genetic Constraint widget displays these observed/expected (oe) values, along with their confidence interval. In the example below, KIT has an oe of 0.17 (0.098 - 0.305) for predicted loss of function variants (pLoF). In other words, 17% of the expected LoF variants were observed, and therefore, KIT is highly likely to be under selection against LoF variants.

The Genetic Constraint widget for KIT, which is classed as “5/5 very high constraint”.

Chemical probes

We have added a new source of chemical probes data: the Probes&Drugs database, which pulls compounds from 10 different sources. The Platform now contains information on 3,769 different compounds, which contribute to the annotation of 624 targets.

Chemical probes data for FGFR3, which now features an assessment of the probe’s quality from Probes&Drugs, additional information on the scoring of each compound, and links out to the original reference used by Probes&Drugs.

Data updates: highlights


The Experimental Factor Ontology (EFO) had a new release in August — 3.34.0 — which includes a major restructuring of the neoplasm branch, using the hierarchy created by Mondo. Users of the Platform can therefore expect to see many changes, such as in the propagation of ontologies.

For example, ovarian squamous cell carcinoma, which was erroneously considered a child term of skin cancer, is now nested under carcinoma and squamous cell neoplasms.


We have updated to ChEMBL 29, resulting in an overall increase in all of their annotations. Consequently, the Platform now integrates data on an additional 393 drugs.

One such compound is Sutimlimab, a promising monoclonal antibody currently under investigation as a treatment for hemolysis in cold agglutinin disease.

For more information on this update, read the full ChEMBL Blog post.

Mouse Genome Informatics

We have integrated the latest data on mouse phenotypes from the Mouse Genome Informatics database. Furthermore, we now only include mouse phenotypes which can be mapped to human phenotypes and diseases displayed in the Platform, in line with Open Targets' focus on human disease. The accompanying widget has also been streamlined.

In case you missed it: AlphaFold data

We announced last month that we had included DeepMind’s AlphaFold protein structure predictions on our target profile pages.

“Predicted structural regions with different degrees of confidence from AlphaFold provide a unique angle to explore unknown mechanisms of disease,” explained Platform Coordinator David Ochoa in a Twitter thread.

Other updates and redesigns

  • The comparative genomics widget has been updated, and now includes information on paralogs. For example, the ACHE target profile page now displays additional information on 6 human paralogs.
  • The safety widget has also been redesigned to better display the key pieces of information, so that users can quickly determine the consequences of modulating the target in question. See for example the AR target profile page.
  • The Open Targets literature pipeline uses natural entity recognition to identify sentence-level matches for target, disease, and drug names. We have improved the processing to better disambiguate synonyms which could refer to multiple entities.

Thoughts, comments, or queries? Let us know on the Open Targets Community!

This post was updated on 01.03.22 to correct the reported number of associations.