Missing data in multi-omics integration: Recent advances through artificial intelligence

Area of Science:

• Bioinformatics
• Computational Biology
• Systems Biology

Background:

• Biological systems are understood through multi-omics data, integrating various biomolecule measurements.
• Challenges in multi-omics integration include heterogeneous data and missing measurements across samples.
• Missing data arises from experimental limitations like cost or instrument sensitivity.

Purpose of the Study:

• To review recent artificial intelligence and statistical learning methods for multi-omics data integration.
• To focus on approaches that specifically handle partially observed samples (missing data).
• To discuss traditional missing data workflows, their limitations, and future directions.

Main Methods:

• Review of recently developed artificial intelligence and statistical learning techniques.
• Analysis of methods designed to handle missing data in multi-omics datasets.
• Comparison of different approaches for dealing with partially observed samples.

Main Results:

• Identified and described recent methods for multi-omics integration that address missing data.
• Highlighted the specific use cases and data-handling strategies of these advanced techniques.
• Provided an overview of traditional missing data workflows and their inherent limitations.

Conclusions:

• Advanced AI and statistical learning methods are crucial for effective multi-omics data integration, especially with missing data.
• Handling missing data is a key challenge, and recent methods offer improved solutions.
• Future research should explore further developments and the generalizability of these solutions beyond multi-omics.