17-Nov-2023
Coordinated by Instruct-ERIC, the Fragment-Screen project aims to develop innovative instrumentation and make significant advances in fragment-based drug discovery (FBDD). The project brings together experts in a variety of structural biology domains, including cryo-EM, magnetic resonance research, X-ray techniques, and mass spectrometry, to find new targets and methods in FBDD.
Fragment-Screen supplements the experimental research being carried out by its academic and industrial partners with artificial intelligence. AI has become a progressively more crucial element of structural biology research, with the capacity to assess thousands of potential proteins, ligands, and drug targets, or to predict protein structures to an increasingly accurate level before experimentation.
The capabilities of AI and machine learning (ML) for the identification of novel discoveries (including drug development) is highlighted by Bustillo et al (2023).
The paper, which cites support from Fragment-Screen, explains the ever-improving capabilities of ML in chemical advances. For example, ML has long been an important tool for determining the predicted structure of proteins based on previously established data, or perhaps identifying potential drug targets for SARS-CoV-2. However, many important scientific discoveries have been brought about through curiosity-driven research, and the team explored the extent to which ML could be used to advance this fruitful form of research.
This includes a higher emphasis on smaller and higher quality datasets (rather than large stores of varying quality data), whilst also including “negative” results. This can help to inform ML tools in a different way and identify potentially totally novel and unexpected results.
This curiosity-driven research is the cornerstone of Fragment-Screen, which ambitiously aims to explore libraries of fragment data to develop a wide range of potential treatments, expanding the current arsenal of drugs and therapeutic offerings.
Figure 1. Curiosity-driven research as a strategy to pursue the ‘unknown’. Research conducted in areas that are densely populated by prior work (orange) serves to consolidate intuition and contributes to incremental gains in knowledge. In contrast, research conducted in sparsely populated areas of the search space (green and blue) pushes the boundaries of knowledge.
The merging of experimental and ML techniques for drug discovery is happening at an increasingly rapid pace – as Bustillo et al write, there may soon come a point where ML can create novel approaches to drug discovery at the same rate and ingenuity. Tiago Rodrigues from FFUL, one of the PIs within Fragment Screen, commented “exploitation of AI algorithms is the norm, but that can lead to the discovery of ‘known knowns’. What we are trying to explore and identify are the ‘unknown unknowns’ in fragment-based drug design. To that end, we are implementing novel AI toolkits together with the team of Teo Laino at IBM.”
The Fragment-Screen project is the latest foil for this game-changing collaboration.
