The ChEMBL-og

One year ago we published a new version of our target prediction models and since then we've been working on its implementation for the upcoming ChEMBL 26 release. What did we do? First of all we re-trained the models with the LightGBM library instead of using scikit-learn. By doing this and tuning a bit the parameters our prediction timing improved by 2 orders of magnitude while keeping comparable prediction power. Having quicker models allowed us to easily implement a simple web service providing real time predictions. Since we are currently migration to a more sustainable Kubernetes infrastructure it made sense to us to directly write the small target prediction web service as a cloud native app. We then decided to give OpenFaaS a try as a platform to deploy machine learning models. OpenFaaS is a framework for building serverless functions with Docker and Kubernetes. It provides templates for deploying functions as REST endpoints in many different programming lang...

The ChEMBL team will be heading off for Christmas soon, but just before we do, we wanted to share some updates... First, thanks to all of our many users and collaborators and we wish you all a happy holiday season and a productive 2020! Thanks also to everyone who helped us celebrate 10 years of ChEMBL at our symposium in October. For those who were unable to make it on the day, many of the talks and posters are available here . Over the last few months we've been busy working on ChEMBL_26, which we plan to release early next year. There will be some important changes in this release: We are now using RDKit  for almost all of our compound-related processing. For the first time in ChEMBL_26, this will include compound standardisation (look out for more info on this in the new year), salt-stripping, generation of canonical smiles, structural alerts, substructure searches and similarity searches (via FPSim2 ). Therefore, all molecules have been reprocessed and you may ...

Two new 'Browse' pages have been added to the interface;  Browse Drug Mechanisms  and  Browse Drug Indications . Users can now access these 2 pages directly to explore all the data. Or alternatively, they can land on these pages from drugs, compounds and targets in ChEMBL. Accessing all the data from the main page The 'circles' visualisation on the main page shows a summary of the entities in ChEMBL. Circles for Drug Mechanisms of Action and Drug Indications have been added. By clicking on the circles, you will be taken to a page that allows you to explore the corresponding entity.  Visualisation that summarises the entities in ChEMBL, Drug Mechanisms of Action and Drug Indications are now included. The Browse Drug Mechanisms and Browse Drug Indications pages allow you to use filters, link to other entities, and download the data in the same way as the other 'Browse' pages. All Drug Mechanism data. All Drug Indication data. Acce...

A new text filter has been added to the  search results  and the ' Browse ' pages of the interface. This filter is shown as a small search bar at the top-right of tables and card pages. It can be used as a simple and fast way to filter a set of items. The filter appends a new query to the current query to match the term entered with all the available fields that are non-numeric. It is based on the Querystring query of Elasticsearch , so wildcards can be used in the search box. To see an example of how it works, you can follow these steps: Go to the  Browse Drugs  page: https://www.ebi.ac.uk/chembl/g/#browse/drugs Use the filters to the left to select only Phase 4 drugs with no Rule of Five violations:  Enter the term '*antibacterial*' on the search box and click on the search button: It will match the term on the following fields: Parent Molecule ChEMBL ID, Synonyms, Research Codes, Applicants, USAN Stem, ATC Codes, USA...

CuPy is a really nice library developed by a Japanese startup and supported by NVIDIA that allows to easily run CUDA code in Python using NumPy arrays as input. It also provides interoperability with Numba (just-in-time Python compiler) and DLPackAt (tensor specification used in PyTorch, the deep learning library). CUDA is a parallel computing platform and application programming interface that allows using GPUs for general purpose, not only graphics related computing. Just to give an idea of the level of parallelization it can be achieved with it, a not very expensive consumer's GPU like the NVIDIA GTX 1080 comes with 2560 CUDA cores. Because at ChEMBL we love anything that makes Python fast and that is well integrated with NumPy we couldn't resist to give it a try! Let's go through a example to see how it is working... Google colab notebook . Colab provides the option to run notebooks in GPU and CuPy is already installed on the default Python environment :) ...

Made with Visme Infographic Maker

  Update: KNIME protocol with the model available thanks to Greg Landrum. Update: New code to train the model and ONNX exported trained models available in github . The use and application of multi-task neural networks is growing rapidly in cheminformatics and drug discovery. Examples can be found in the following publications: - Deep Learning as an Opportunity in VirtualScreening - Massively Multitask Networks for Drug Discovery - Beyond the hype: deep neural networks outperform established methods using a ChEMBL bioactivity benchmark set But what is a multi-task neural network? In short, it's a kind of neural network architecture that can optimise multiple classification/regression problems at the same time while taking advantage of their shared description. This blogpost gives a great overview of their architecture. All networks in references above implement the hard parameter sharing approach. So, having a set of activities relating targets and molecules we can tra...