SAMPL7 protein-ligand challenge: A community-wide evaluation of computational methods against fragment screening and pose-prediction

Journal article

The work presented in this thesis focuses on the use of molecular dynamics (MD) and enhanced sampling methods for investigating ligand binding poses and determining protein-ligand binding affinity. Good pre diction of the arrangement of these complexes and their strength is crucial for successful structure based drug design (SBDD) efforts so this thesis makes a significant contribution in furthering the use of computational tools in SBDD. First, chapter 3 presents OpenBPMD, an open-source Python re implementation of binding pose metadynamics (BPMD), a MD-based tool for ranking ligand poses from a set of candidates derived from dock ing. The role of accurate water positioning on the performance of the algorithm is also investigated, showed how the combination with a grand canonical Monte Carlo algorithm improves the accuracy of the predic tions. Then chapter 4 explains how the funnel metadynamics (fun-metaD) algorithm was implemented on a high-performance MD engine, OpenMM. This implementation was validated on host-guest systems. Afterwards a larger data set is interrogated, examining the effects on host-guest bind ing by varying the water model (TIP3P, OPC3 and OPC) and the partial atomic charge assignment methods, AM1-BCC and RESP. Finally, chapter 5 investigates the binding of fragment-like ligands in three different protein targets by applying fun-metaD. Advancements are made on funnel-shaped restraint automation and a new set of collective variables (CV) is tested as well. However, a lack of convergence due to an excess of metadynamic bias and missing slow degrees of freedom is observed. In order to address these issues, chapter 6 delves into apply ing a neural network-based CV, called Deep-LDA, and a novel enhanced sampling algorithm, termed on-the-fly probability-enhanced sampling. Although smooth converging, some issues in pose discrimination still re main

Authors: Grosjean, H; Işık, M; Aimon, A; Mobley, D; Chodera, J

• Publication status: Published
• Peer review status: Peer reviewed
• Publisher: Springer
• Journal: Journal of Computer-Aided Molecular Design
• Volume: 36
• Issue: 4
• Pages: 291-311
• Publication date: 2022-04-15
• DOI: 10.1007/s10822-022-00452-7
• EISSN: 1573-4951
• ISSN: 0920-654X
• Language: English
• Keywords: Biology; Computational biology; Workflow; Bromodomain; Fragment (logic); Algorithm; Biochemistry; Drug discovery; Computer science; Chemistry