DrugAppy - An end-to-end deep learning framework for computational drug discovery.

Area of Science:

• Oncology
• Computational Chemistry
• Drug Discovery

Background:

• Targeting oncogenic vulnerabilities is key for advanced cancer treatment, but drug development is costly and slow.
• Computational tools offer a promising strategy to accelerate and optimize the drug design process.
• Existing methods require significant time and resources for identifying effective anti-cancer agents.

Purpose of the Study:

• To introduce DrugAppy, a hybrid computational tool for identifying novel cancer drug inhibitors.
• To leverage Artificial Intelligence (AI) and computational chemistry for efficient drug discovery.
• To validate the tool's efficacy through case studies on PARP and TEAD protein families.

Main Methods:

• DrugAppy integrates AI algorithms with computational and medicinal chemistry techniques.
• Utilizes High Throughput Virtual Screening (HTVS) with SMINA and GNINA, and Molecular Dynamics (MD) with GROMACS.
• Incorporates AI models for predicting drug pharmacokinetics, selectivity, and activity.

Main Results:

• DrugAppy successfully identified novel inhibitors for Poly(ADP-ribose) polymerase (PARP) and the transcriptional enhanced associate domain (TEAD) family.
• Two molecules targeting PARP1 showed activity comparable to the existing drug olaparib.
• One identified compound for TEAD4 outperformed the reference inhibitor IK-930 in vitro.
• Confirmed target engagement for identified active compounds.

Conclusions:

• The DrugAppy workflow effectively discovers novel molecular structures with significant therapeutic potential.
• The tool demonstrates a viable approach to accelerate the identification of potent anti-cancer agents.
• This methodology provides a powerful platform for targeting key proteins in cancer research.