Advances in predicting omics profiles from imaging data.

Area of Science:

• Biomedical Imaging
• Computational Biology
• Genomics

Background:

• Traditional molecular profiling is complex and costly.
• Predicting molecular data directly from medical images is an emerging, cost-effective alternative.
• Existing reviews focus on specific biomarkers or diseases, lacking a comprehensive overview.

Purpose of the Study:

• To provide a comprehensive review of methods predicting molecular 'omics' data from imaging.
• To cover DNA aberrations, bulk, single-cell, and spatial transcriptomics.
• To explore diverse disease contexts and imaging modalities.

Main Methods:

• Review of studies employing deep learning for image processing, feature extraction, and molecular prediction.
• Analysis of modern statistical frameworks for image-based omics prediction.
• Inclusion of diverse imaging modalities and omics data types.

Main Results:

• Deep learning strategies are widely used for image processing, feature extraction, aggregation, and molecular prediction.
• Diverse applications of deep learning and statistical frameworks for image-based omics prediction are highlighted.
• Significant progress in predicting various omics data (DNA, transcriptomics) from imaging.

Conclusions:

• Image-based omics prediction is a promising field with broad clinical relevance.
• Inferred molecular data improves understanding of molecular-visual feature relationships.
• This approach paves the way for novel diagnostic and therapeutic applications.