The promise of explainable deep learning for omics data analysis: Adding new discovery tools to AI

Area of Science:

• Bioinformatics
• Computational Biology
• Artificial Intelligence in Life Sciences

Background:

• Deep learning (DL) excels at processing large, heterogeneous datasets, significantly impacting omics data analysis.
• Understanding complex biological systems is a key challenge, amplified by the growing volume of omics data.
• Current DL methods often lack transparency, posing challenges for critical applications like clinical environments.

Approach:

• This review explores the transformative impact of explainable deep learning (EDL) on biological data interpretation.
• EDL provides transparency and new insights, moving beyond prediction and classification.
• The review highlights EDL's application across diverse sectors, including genomics, radiomics, drug design, and clinical trials.

Key Points:

• Explainable deep learning (EDL) enhances transparency and interpretability in biological data analysis.
• EDL facilitates the discovery of novel biological insights from complex datasets.
• Applications span genome engineering, genomics, radiomics, drug design, and clinical trials.

Conclusions:

• Explainable deep learning (EDL) offers a powerful paradigm shift for interpreting biological data.
• Life scientists are encouraged to adopt EDL tools for deeper understanding and discovery.
• Resources are suggested to facilitate the initial adoption of EDL in research.