Early-stage drug discovery has been limited by initial hit identification and lead optimization and their associated costs. Ultra-large virtual screens (ULVSs), which involve the virtual evaluation of massive numbers of molecules to engage a macromolecular target, have the ability to significantly alleviate these problems, as was recently demonstrated in multiple studies. Despite their potential, ULVSs have so far only explored a tiny fraction of the chemical space and of available docking programs. Here, we present VirtualFlow 2.0, the next generation of the first open-source drug discovery platform dedicated to ultra-large virtual screenings. VirtualFlow 2.0 provides the REAL Space from Enamine containing 69 billion drug-like molecules in a "ready-to-dock" format, the largest library of its kind available to date. We provide an 18-dimensional matrix for intuitive exploration of the library through a web interface, where each dimension corresponds to a molecular property of the ligands. Additionally, VirtualFlow 2.0 supports multiple techniques that dramatically reduce computational costs, including a new method called Adaptive Target-Guided Virtual Screening (ATG-VS). By sampling a representative sparse version of the library, ATG-VS identifies the sections of the ultra-large chemical space that harbors the highest potential to engage the target site, leading to substantially reduced computational costs by up to a factor of 1000. In addition, VirtualFlow 2.0 supports the latest deep learning- and GPU-based docking methods, allowing further speed-ups by up to two orders of magnitude. VirtualFlow 2.0 supports 1500 unique docking methods providing target-specific and consensus docking options to increase accuracy and has the ability to screen new types of ligands (such as peptides) and target receptors (including RNA and DNA). Moreover, VirtualFlow 2.0 has many advanced new features, such as enhanced AI and cloud support. We demonstrate a perfectly linear scaling behavior up to 5.6 million CPUs in the AWS Cloud, a new global record for parallel cloud computing. Due to its open-source nature and versatility, we expect that VirtualFlow 2.0 will play a key role in the future of early-stage drug discovery.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.