Benchmarking omics-based prediction of asthma development in children

Area of Science:

• Computational biology
• Genomics
• Systems biology

Background:

• Asthma is a complex, heterogeneous disease with significant morbidity.
• High-throughput multi-omics data offer layered molecular insights into complex diseases.
• A systematic comparison of computational methods for omics-based asthma prediction is needed.

Purpose of the Study:

• To investigate and benchmark computational methods for disease status prediction using multi-omics data.
• To evaluate the performance of various omics data combinations for asthma prediction.

Main Methods:

• Systematic benchmarking of 18 computational methods.
• Utilized 63 combinations of six omics data types (GWAS, miRNA, mRNA, microbiome, metabolome, DNA methylation) from the VDAART cohort.
• Performance evaluation using standard metrics for each omics combination.

Main Results:

• Logistic Regression, Multi-Layer Perceptron, and MOGONET demonstrated superior performance.
• The combination of transcriptional, genomic, and microbiome data yielded the best prediction accuracy.
• Inclusion of clinical data improved prediction for some, but not all, omics combinations.

Conclusions:

• Specific multi-omics data combinations can optimize asthma development prediction in children.
• Certain computational methods significantly outperform others in multi-omics asthma prediction.