Categorical Matrix Completion With Active Learning for High-Throughput Screening.

Area of Science:

• Computational Biology
• Biotechnology
• Drug Discovery

Background:

• Wet-lab automation facilitates high-throughput experiments, but exhaustive screening is often suboptimal and costly.
• Current high-throughput screening (HTS) methods may not be cost-effective due to exploring all possibilities.

Purpose of the Study:

• To design an open-source artificial intelligence (AI) model for guiding high-throughput screening experiments.
• To improve the cost-effectiveness and efficiency of HTS by prioritizing informative experiments.

Main Methods:

• Developed an AI model integrating categorical matrix completion and active machine learning.
• Focused on screening chemical compound effects on protein sub-cellular locations.
• Incorporated innovations like margin sampling for uncertainty estimation and prioritized exploration over exploitation.

Main Results:

• The model demonstrated robustness across diverse scenarios in simulations.
• Real-world data validated the model's applicability in wet-lab high-throughput screening.
• The model's exploration ability was key to its success, as shown by matrix rank and experiment coverage analyses.

Conclusions:

• The proposed AI model offers a more optimal and cost-effective approach to high-throughput screening.
• The model's design, emphasizing exploration, enhances the efficiency of identifying chemical compound effects.
• The open-source model is applicable to real-world wet-lab experiments, advancing automated scientific discovery.