An open-source screening platform accelerates discovery of drug combinations.

Area of Science:

• Pharmacology
• Biotechnology
• Computational Biology

Background:

• Drug combination screening is crucial but limited by experimental complexity and low throughput.
• Existing liquid handling systems lack integration with combination-specific analytical methods.
• Accelerating the discovery of effective drug combinations is a key challenge in modern medicine.

Purpose of the Study:

• To introduce Combocat, an open-source framework for ultrahigh-throughput drug combination screening.
• To integrate acoustic liquid handling with machine learning for predictive inference of drug effects.
• To reduce the number of experiments required for comprehensive drug combination analysis.

Main Methods:

• Developed Combocat, a framework combining acoustic liquid handling and machine learning-based inference.
• Generated a reference dataset of over 800 drug combinations in a 10x10 matrix format across multiple cell types.
• Trained a predictive model using the generated dataset to infer drug combination effects from sparse data.

Main Results:

• Achieved ultrahigh-throughput drug combination screening, enabling the testing of 9,045 combinations in a neuroblastoma cell line.
• Demonstrated accurate inference of drug combination effects from sparse data, drastically reducing experimental measurements.
• Established a new benchmark for the largest number of drug combinations screened in a single cell line.

Conclusions:

• Combocat offers a scalable solution for accelerating the discovery of novel drug combinations.
• The integration of advanced dispensing technologies and predictive computational modeling enhances screening efficiency.
• This approach significantly reduces the resources needed for extensive drug combination studies.