Progress and new challenges in image-based profiling.

Area of Science:

• Computational Biology
• Bioinformatics
• Cellular Imaging

Background:

• Image-based profiling has transformed cell phenotype analysis for over 20 years.
• It generates high-throughput, unbiased measurements from microscopy data, aiding drug discovery and genomics.
• Deep learning has significantly advanced computational approaches in this domain.

Purpose of the Study:

• To review the computational evolution of image-based profiling.
• To detail bioinformatics processes from feature extraction to batch correction.
• To provide a roadmap for researchers navigating progress and challenges.

Main Methods:

• Review of computational landscape and bioinformatics processes.
• Discussion of deep learning's impact on feature extraction and analysis.
• Examination of advancements in single-cell analysis, similarity metrics, and new modalities.

Main Results:

• Deep learning has reshaped image-based profiling methodologies.
• Advancements include single-cell analysis, new similarity metrics, and expanded modalities (e.g., 3D organoids, temporal imaging).
• Growth in public benchmarks and open-source software enhances reproducibility and collaboration.

Conclusions:

• The field has evolved significantly with computational and deep learning advancements.
• Substantial challenges remain, especially for 3D/temporal data, quality control, and feature interpretation.
• Continued development is crucial for addressing emerging data modalities and ensuring robust analysis.