TL;DR
Researchers have developed a machine learning approach to identify small molecules that bind to SARS-CoV-2 Mpro, a key viral enzyme. While these molecules didn't inhibit the enzyme's activity, they bound to an unexpected allosteric site, paving the way for designing novel PROTAC-based therapeutics that degrade the virus's protein.
What Problem Does This Paper Address?
The COVID-19 pandemic highlighted the urgent need for effective antiviral treatments. SARS-CoV-2's main protease (Mpro) is a critical target for drug development due to its essential role in viral replication and its conservation across coronaviruses. While many efforts have focused on developing inhibitors that directly block Mpro's active site, this paper explores an alternative strategy: identifying molecules that bind to allosteric sites. Allosteric inhibitors can offer advantages like reduced resistance and potentially better safety profiles. The challenge lies in discovering these less-obvious binding sites and molecules that interact with them, especially when direct inhibition isn't observed.
Methods
The study employed a multi-pronged approach:
- Machine Learning (ML) for Virtual Screening: A Support Vector Machine (SVM) model, previously trained on a dataset of known Mpro inhibitors, was used to screen a large library of commercially available compounds. This model was designed to predict potential binders based on physicochemical properties rather than structural similarity.
- Microscale Thermophoresis (MST): This biophysical technique was used to experimentally validate the ML predictions. MST measures the movement of molecules in response to a temperature gradient, allowing for the quantification of binding affinities (Ka values) between small molecules and the Mpro protein.
- Enzymatic Assays: Standard assays were performed to determine if the identified binders inhibited Mpro's enzymatic activity.
- Computational Studies:
- SiteMap: This tool was used to identify potential binding pockets on the Mpro protein structure, including both known active sites and less-explored accessory sites.
- Molecular Docking: This technique predicted the binding poses of the identified molecules within the identified pockets.
- Molecular Dynamics (MD) Simulations: Long-term simulations were conducted to assess the stability of the ligand-protein interactions and to gain insights into the dynamic behavior of the complexes.
Datasets
- PostEra COVID-19 Moonshot Dataset: Used for training the initial SVM model. This dataset contains small molecules annotated with experimentally measured IC50 values.
- Commercial Compound Libraries: Approximately two million compounds from vendors like MolPort, Asinex, and ChEMBL were screened using the trained SVM model.
- PDB Accession Code: 6W63: The crystal structure of SARS-CoV-2 Mpro was used for computational studies.
Key Insights and Ideas
- Allosteric Binding is Key: The most significant finding is that the identified small molecule binders, while showing affinity for Mpro, did not inhibit its enzymatic activity. This strongly suggested they were not binding to the active site.
- Discovery of an Accessory Pocket: Computational analysis successfully identified an understudied allosteric pocket on the Mpro protein, distinct from the catalytic site. This pocket proved to be a stable binding site for the identified compounds.
- PROTACs as a Novel Therapeutic Avenue: The binding to an allosteric site, rather than the active site, makes these compounds ideal starting points for designing PROTACs. PROTACs are bifunctional molecules that recruit cellular machinery to degrade target proteins, offering a distinct mechanism of action compared to traditional inhibitors.
- Synergy of ML and Biophysics: The study demonstrates the power of combining ML-driven virtual screening with experimental biophysical validation (MST) to identify novel binders, even when they don't exhibit the expected inhibitory activity.
Key Results
- 180 Compounds Selected for Validation: The SVM model screened millions of compounds, prioritizing 180 for experimental testing.
- Three Promising Hits Identified: MST experiments identified three compounds with binding affinities in the low micromolar range.
- Compound 7 Shows High Affinity: Compound 7 exhibited the highest affinity, with a Ka value of 2.8 ± 0.9 µM.
- No Enzymatic Inhibition: None of the identified binders showed significant inhibition of Mpro activity at a 20 µM concentration.
- Stable Allosteric Binding: MD simulations confirmed that compounds 7, 8, and 9 remained stably bound within the identified allosteric pocket throughout the simulation period.
- Rational Design for PROTACs: The study proposed a retrosynthetic pathway for designing PROTACs based on compound 7, highlighting a specific position for linker attachment and outlining synthetic steps.
Limitations and Future Work
- Limited Enzymatic Inhibition: The primary limitation is the lack of direct enzymatic inhibition by the identified compounds. While this points to allosteric binding, it means these molecules themselves are not direct Mpro inhibitors.
- Allosteric Site Functionality: The paper acknowledges that the precise functional role of the identified allosteric site and how binding to it influences Mpro's activity remains to be fully elucidated.
- PROTAC Efficacy Not Yet Tested: The study proposes PROTAC designs but does not present experimental data on their efficacy in degrading Mpro or their antiviral activity.
- Future Work: The authors suggest further experimental validation of the proposed PROTAC designs and a deeper investigation into the functional consequences of allosteric site binding. They also emphasize the potential for this approach to discover binders for other viral targets.
Summary updated