Comparison of multivariate data analysis strategies for high-content screening

Area of Science:

• Biomedical research
• Cell biology
• Data science

Background:

• High-content screening (HCS) generates complex, multivariate single-cell data.
• Data preprocessing, including normalization and dimensionality reduction, is crucial for extracting meaningful information from HCS datasets.
• No published comparisons exist for different HCS data processing strategies.

Purpose of the Study:

• To comparatively evaluate unbiased methods for dimensionality reduction and cell population summarization in HCS.
• To assess the impact of different data processing strategies on the analysis of HCS data.

Main Methods:

• Comparative evaluation of unbiased dimensionality reduction techniques.
• Assessment of cell population summarization methods, including percentile values at the well level.
• Monitoring prediction accuracies and Z' factors of control compounds using a cell cycle HCS dataset.

Main Results:

• Dimensionality reduction generally decreased the discrimination between control samples.
• Summarizing cell populations at the well level using percentile values yielded high classification accuracy.
• The study identified effective strategies for analyzing multiparametric HCS data.

Conclusions:

• A systematic data analysis pipeline can facilitate the review of alternative HCS data processing strategies.
• Well-level summarization using percentile values is an effective approach for HCS data analysis.
• The findings provide a framework for optimizing HCS data interpretation in biomedical research.