Benchmarking AlphaFold ‐enabled molecular docking predictions for antibiotic discovery

Abstract

Efficient identification of drug mechanisms of action remains a challenge. Computational docking approaches have been widely used to predict drug binding targets; yet, such approaches depend on existing protein structures, and accurate structural predictions have only recently become available from AlphaFold2. Here, we combine AlphaFold2 with molecular docking simulations to predict protein-ligand interactions between 296 proteins spanning Escherichia coli's essential proteome, and 218 active antibacterial compounds and 100 inactive compounds, respectively, pointing to widespread compound and protein promiscuity. We benchmark model performance by measuring enzymatic activity for 12 essential proteins treated with each antibacterial compound. We confirm extensive promiscuity, but find that the average area under the receiver operating characteristic curve (auROC) is 0.48, indicating weak model performance. We demonstrate that rescoring of docking poses using machine learning-based approaches improves model performance, resulting in average auROCs as large as 0.63, and that ensembles of rescoring functions improve prediction accuracy and the ratio of true-positive rate to false-positive rate. This work indicates that advances in modeling protein-ligand interactions, particularly using machine learning-based approaches, are needed to better harness AlphaFold2 for drug discovery.

Protein engineering of amine transaminases

“…Other methodologies, including algorithms for stability prediction and definition of search space for mutagenesis, can be refined with the help of AlphaFoldproduced structures (Zhang Y. et al, 2021). A recent report indicates that an AlphaFold-generated structure can be of sufficient quality to allow structure-based design since the predicted protein-ligand complexes were indistinguishable from their experimentally-determined counterparts (Wong et al, 2022). In June 2022, the AlphaFold team released 200 million protein structures (https://alphafold.ebi.ac.uk/), which essentially covers all catalogued proteins, significantly increasing the number of protein structures available.…”

Beyond sequence: Structure-based machine learning

“…In these contexts, ML approaches are used to improve scoring functions of binding affinity and plausible docking poses [121] , [116] , [81] , [138] , [139] . Indeed, [140] show that computationally predicted structures perform on par with experimental structures at reverse docking tasks - although the docking and scoring methods themselves could use major improvements to further drug discovery and design.…”

Comprehensive Survey of Consensus Docking for High-Throughput Virtual Screening

“…Wong et al used 12 essential proteins, 218 active compounds, and 100 inactive compounds to predict antibacterial inhibitors and found that, although models had low performance, the use of rescoring strategies may have acceptable predictive power for certain proteins. They concluded that the limitations in benchmarking are not due to the AlphaFold structures itself, but to the methods to accurately model the protein-ligand interactions [27]. Other studies have identified potential inhibitors of WD40 repeat and SOCS box containing 1 protein (WSB1), a clinically relevant drug target, by means of AlphaFold and virtual screening approaches [28].…”

“…They found that rescoring based on simplistic knowledgebased scoring functions, e.g., measuring interaction fingerprints, appears to outperform modern machine learning methods, highlighting the importance of the use of rescoring methods to properly detect the most potent binders [103]. Similarly, recent studies have demonstrated that the use of machine learning approaches to rescore docking poses greatly enhances the performance of structural models and that ensembles of rescoring functions increase prediction accuracy [27]. They concluded that the use of empirical data to assess docking predictions is a key factor to improve the prediction of protein-ligand interaction in drug discovery.…”