Active Learning Strategies for Phenotypic Profiling of High-Content Screens

Area of Science:

• Biotechnology
• Computational Biology
• Drug Discovery

Background:

• High-content screening (HCS) is crucial for drug discovery and biological research.
• Supervised machine learning (SML) automates phenotypic classification in HCS.
• Labeling data for SML models is time-consuming and requires expert input.

Purpose of the Study:

• To investigate the impact of active learning (AL) on single-cell phenotype recognition in HCS.
• To evaluate combinations of AL strategies and SML methods.
• To determine if AL can reduce the cost of obtaining labeled data while maintaining classification performance.

Main Methods:

• Utilized data from three large-scale RNA interference HCS experiments.
• Applied various active learning strategies in conjunction with popular SML methods.
• Assessed performance based on phenotypic target identification and accuracy.

Main Results:

• Active learning significantly reduced the time investment required for model training.
• AL strategies successfully identified the same phenotypic targets as traditional SML.
• Certain combinations of AL strategies and SML methods demonstrated superior performance.

Conclusions:

• Active learning is an effective strategy to overcome data labeling bottlenecks in HCS.
• AL minimizes expert time and cost while achieving high accuracy in phenotype recognition.
• Optimized AL and SML combinations can enhance the efficiency of HCS data analysis.