Machine learning and molecular simulation-based discovery of novel RIPK1 inhibitors

Area of Science:

• Medicinal Chemistry
• Computational Drug Discovery
• Biochemistry

Background:

• Receptor-interacting serine/threonine-protein kinase 1 (RIPK1) is a key regulator of necroptosis and implicated in neurological disorders, inflammation, and cancer.
• While RIPK1 is a druggable target, no approved therapeutics currently exist, highlighting the need for novel inhibitors.

Purpose of the Study:

• To identify novel RIPK1 inhibitors from marine natural products using computational methods.
• To evaluate the binding affinity, pharmacokinetic properties, and stability of identified compounds with RIPK1.

Main Methods:

• A machine-learning model (Gradient Boosting) was trained on ChEMBL data to screen the Comprehensive Marine Natural Product Database (CMNPD).
• ADMET filtration, molecular docking, molecular dynamics simulations, and MM-PBSA calculations were employed to assess potential inhibitors.
• Four compounds (CMNPD23788, CMNPD14579, CMNPD15831, CMNPD26709) were identified and characterized computationally.

Main Results:

• The machine-learning model demonstrated high predictive performance (accuracy=0.97, F1=0.97, ROC-AUC=0.99).
• Four marine natural products exhibited high binding affinities to the RIPK1 kinase domain, comparable to existing inhibitors.
• Two compounds showed predicted CNS penetrance, and two others predicted oral bioavailability similar to a reference inhibitor. Molecular dynamics simulations confirmed complex stability.

Conclusions:

• CMNPD23788, CMNPD14579, CMNPD15831, and CMNPD26709 are computationally identified as potential RIPK1 inhibitors.
• These marine-derived compounds possess favorable predicted pharmacokinetic properties, including CNS penetration and oral bioavailability.
• Further experimental validation is necessary to confirm the therapeutic potential of these compounds in necroptosis-related diseases.