Multi-omics integration for neuroblastoma clinical endpoint prediction

Area of Science:

• Bioinformatics
• Computational Biology
• Cancer Genomics

Background:

• High-throughput omics technologies generate complex cancer data.
• Integrating multi-omics data offers deeper insights into cancer biology.
• Personalized cancer therapies rely on comprehensive data analysis.

Purpose of the Study:

• To apply Integrative Network Fusion (INF) for neuroblastoma patient outcome prediction.
• To integrate RNA-Seq, microarray, and array comparative genomic hybridization data.
• To explore autoencoders for integrating microarray expression and copy number data.

Main Methods:

• Utilized Integrative Network Fusion (INF), a framework combining similarity network fusion and machine learning.
• Applied INF to predict neuroblastoma patient outcomes using multi-omics data.
• Employed autoencoders for integrating specific omics data types.

Main Results:

• INF effectively integrated multiple data sources for patient classification.
• The autoencoder approach showed promising results in improving survival endpoint classification.
• Autoencoders facilitated the discovery of patient groups with distinct overall survival (OS) curves.

Conclusions:

• INF is an effective method for multi-source data integration in cancer research.
• Autoencoder-based latent space representation aids in patient stratification and outcome prediction.
• This study highlights the potential of integrated omics data for personalized neuroblastoma treatment strategies.