Artificial intelligence in cancer research: learning at different levels of data granularity

Area of Science:

• Oncology and Computational Biology
• Artificial Intelligence in Medicine
• Big Data Analytics in Healthcare

Background:

• Cancer research generates diverse data, from genomic to clinical records, creating a big data environment.
• This data landscape is heterogeneous, featuring a mix of big and small datasets with varying characteristics (sample size, labels, types).
• Integrating these diverse data granularities is crucial for a systems-level understanding of cancer.

Purpose of the Study:

• To review the challenges and limitations of applying Artificial Intelligence (AI) to heterogeneous data granularity in cancer research.
• To explore current solutions and advancements in AI for cancer data integration.
• To highlight the importance of interoperability and the synergy between discriminative and generative AI models.

Main Methods:

• Literature review of AI applications in cancer research focusing on data granularity.
• Analysis of challenges posed by diverse data types, sample sizes, and labels.
• Discussion of AI techniques, including discriminative and generative models, for data integration.

Main Results:

• AI integration faces hurdles due to the coexistence of big and small data resources in cancer research.
• Advancing AI interoperability is essential to address varied data descriptors and levels of granularity.
• Synergistic use of discriminative and generative models offers promising solutions for complex cancer data.

Conclusions:

• Addressing data granularity heterogeneity is key for advancing AI in cancer research.
• Enhanced interoperability and hybrid AI models are crucial for a comprehensive systems view of cancer.
• Future directions involve leveraging AI to bridge data gaps and improve cancer understanding through integrated analysis.