Progress and developments of artificial intelligence for drug design
- Marco De Vivo (Istituto Italiano di Technologia, Italy)
- György M Keseru (Research Center for Natural Sciences, Hungary)
- Jonas Boström (AstraZeneca, Sweden)
- Darren Green (GlaxoSmithKline, United Kingdom)
After decades of incremental innovations in drug design, artificial intelligence (AI) tools have started to show promise revolutionizing the way finding new drug targets, identifying viable chemical starting points, designing efficient synthesis routes and defining novel therapeutic markers for translational studies.
The new concept and the associated technologies have the potential alleviating the numbers game in drug discovery as well as making pharma R&D more efficient by automation In early discovery AI can support finding better drugs against well characterized targets by the analysis and prediction of druggable binding sites and exploitable interactions, as well as for a target deconvolution/identification. Deep learning algorithms can transform screening programs by being taught to design optimized compounds and increase hit rates utilizing the knowledge encoded in large published and proprietary datasets. Further applications involve repositioning and redesign programs that search in the known set of drugs to find pharmacotherapies for new indications.
Medicinal chemistry teams realizes multiple benefits from these tools. In drug design, AI technologies are perfectly suited for solving multidimensional optimization problems that is typically the case in drug discovery settings. In synthetic chemistry, deep learning can suggest viable alternatives how to best synthesize the designed compounds. Optimization of chemical reactions using contributes to provide the best compound at the best time that might deliver lower attrition rates as compounds move through the pharma value chain. The efficiency of these processes are often enhanced by automated technologies in both synthesis and testing that provides a previously unseen amount of data feeding and improving AI-based design technologies. Finally, AI technologies have significant impact on clinical research. Its advancement to the clinical trials results the process faster, cutting costs, improving trial quality, and reducing trial times. Finding disease biomarkers and gene signatures support recruiting eligible clinical trial patients and opens new perspectives in personalized therapies. The present workshop covers the key areas of AI driven drug discovery including the theory and practice in target identification, hit finding, drug design and synthetic chemistry.
1. Ching et al, Opportunities and Obstacles for deep learning in biology and medicine, J. R. Soc. Interface 15:20170387
2. Segler et al, Planning chemical syntheses with deep neural networks and symbolic AI, Nature, 555 p604, 2018
3. Schwaller et al, “Found in Translation”: Predicting Outcomes of Complex Organic Chemistry Reactions using Neural Sequence-to-Sequence Model, arXiv:1711.04810
4. Merk et al, De Novo Design of Bioactive Small Molecules by Artificial Intelligence, Mol. Inf., 37, 1700153, 2018
5. Pant et al, Design, Synthesis, and Testing of Potent, Selective Hepsin Inhibitors via Application of an Automated Closed-Loop Optimization Platform, J. Med. Chem., 61, p4335, 2018
6. Gómez-Bombarelli et al, Automatic Chemical Design Using a Data-Driven Continuous Representation of Molecules, ACS Cent. Sci., 4, p268, 2018
7. Segler et al, Generating Focussed Molecule Libraries for Drug Discovery with Recurrent Neural Networks, arXiv:1701.01329
8. Altae-Tran et al, Low Data Drug Discovery with One-Shot Learning, ACS Cent. Sci., 3, p283, 2017