Vendor Workshops

Monday, June 27 | 12:20-1:05PM
Vendor Workshop by Schrödinger

Augmenting Medicinal Chemistry with Digital Chemistry: A CDK2 Inhibitor Design Challenge

Presented by: Wade Miller (Senior Scientist I)

This hands-on design challenge will give participants the opportunity to design, computationally assess, and prioritise novel CDK2 inhibitors. As part of the design challenge, participants will have access to:

• A validated docking and design model
• Predicted physicochemical properties
• Machine learning models for various ADMET properties
• A tool which searches ChEMBL and vendor databases (with over 1 billion total compounds) at rapid speeds to estimate the novelty of designed compounds
• A Target Product Profile MPO that will be used to determine the ‘winning’ design

Monday June 27 | 4:45-5:30PM
Vendor Workshop by Iktos

AI Tools for Medicinal Chemists: Novel Compound Design and Retrosynthesis Prediction

Presented by: Rohit Arora & Brian Atwood (Application Scientists) 

Multi-parameter optimization (MPO) is a major challenge confronting the drug discovery process, making it difficult to identify promising synthetically accessible molecules meeting the Target Product Profile (TPP). To overcome this, we have developed an AI-driven pipeline (Makya and Spaya) for generative de novo drug design to enhance chemical space exploration while tackling the MPO challenge under synthetic constraints. We will discuss how scientists can use these tools in their own projects and how the tools have impacted real-life drug discovery projects.

Wednesday June 29| 12:20-1:05PM
Vendor Workshop by Aseda Sciences

Drug Discovery Platforms, the Predictive Information Gap, and How to Bridge it

Presented by: Andrew A. Bieberich, PhD

Over the past two decades, web-based, GUI-operated platforms have emerged that generally claim to increase drug discovery/development efficiency and enable more rational design processes that do not "repeat the sins of the past".  However, clinical trials still exhibit an approximately 90% overall failure rate, with up to 40% of safety-related failures arising from signals that were not detected during preclinical work.

We suspect that existing computational drug development platforms have failed to change this scenario because they have avoided doing the one thing that is difficult during platform development, and hence expensive and time consuming, which is data transformation into actionable information. One platform model is to provide users with cloud-based data storage and project administration features, coupled with standard data analysis and plotting/visualization tools.  However, creating information with true predictive power (the hard part) is offloaded to the user.

A second platform model is to provide a searchable database of information describing known pharmaceutical molecular structure space, with varying effort expended on curation for accuracy and sophistication of query tools. However, determining which parameters in the curated information may inform potential success of proprietary molecules (again, the hard part) is still ultimately the user’s responsibility.  In either model, the user has no empirical means of directly comparing related biological and chemical features of proprietary molecules with those same features in known pharmaceutical space.

Ten years ago, AsedaSciences began developing a platform to address this challenge that uses carefully curated, pre-existing information, for known pharmaceutical space, coupled with cell-based phenotypic screening and supervised Machine Learning (ML), to directly compare new molecules with thousands of compounds from that known space. This enables users to create predictive information in a semi-guided fashion. We describe how the 3RnD® platform and SYSTEMETRIC® Cell Health Screen address this challenge with design strategy and case studies.