Identification of STAT3 phosphorylation inhibitors using generative deep learning, virtual screening, molecular...

Area of Science:

• Oncology
• Pharmacology
• Computational Chemistry

Background:

• Signal Transducer and Activator of Transcription 3 (STAT3) is a key target for non-small cell lung cancer (NSCLC) therapy.
• Developing effective STAT3 phosphorylation inhibitors is crucial for advancing NSCLC treatment strategies.

Purpose of the Study:

• To discover novel, potent STAT3 inhibitors for NSCLC using a generative deep learning model.
• To identify and validate lead compounds with strong binding affinities and therapeutic potential.

Main Methods:

• Utilized transfer learning and virtual screening to build a generative model for STAT3 inhibitors.
• Employed molecular docking and molecular dynamics (MD) simulations for compound prioritization and validation.
• Assessed STAT3 phosphorylation suppression and nuclear translocation inhibition in cellular models.

Main Results:

• Generated a diverse library of potential STAT3 inhibitors, identifying HG110 as a potent candidate.
• HG110 demonstrated significant suppression of STAT3 phosphorylation at Tyr705 and inhibited nuclear translocation.
• MD simulations confirmed stable binding conformations and favorable interactions for HG106 and HG110, outperforming existing inhibitors.

Conclusions:

• Generative deep learning accelerates the discovery of selective STAT3 inhibitors.
• The identified compounds, particularly HG110, offer a promising therapeutic avenue for NSCLC treatment.
• This approach provides a robust platform for developing targeted cancer therapies.