Accelerated hit identification with target evaluation, deep learning and automated labs: prospective validation in...

Area of Science:

• Drug Discovery and Development
• Computational Chemistry
• Biotechnology

Background:

• Biomedical knowledge analysis, AI virtual screening, and robotic labs can transform drug discovery.
• Prospective studies evaluating integrated approaches are scarce.

Purpose of the Study:

• To prospectively validate an integrated platform for target identification and hit discovery.
• To evaluate the efficacy of AI-driven virtual screening and robotic cloud labs in early-stage drug discovery.

Main Methods:

• Integration of in-house target evaluation (SpectraView) and deep-learning virtual screening (HydraScreen) tools.
• Utilized an automated robotic cloud lab for ultra-high-throughput screening.
• Employed SpectraView to select IRAK1 as the target and prospectively validated a structure-based deep learning model.

Main Results:

• The AI model identified 23.8% of all IRAK1 hits within the top 1% of ranked compounds, outperforming traditional methods.
• Identified three potent (nanomolar) scaffolds for IRAK1, with two representing novel candidates.
• Demonstrated the reliability and robustness of HydraScreen in virtual and high-throughput screening.

Conclusions:

• SpectraView and HydraScreen significantly accelerate target identification and hit discovery.
• AI-driven virtual screening with HydraScreen offers high hit discovery rates and reduces experimental costs for IRAK1.
• The integrated platform expedites early-stage drug discovery processes.