Identification of cancer risk groups through multi-omics integration using autoencoder and tensor analysis.

Area of Science:

• Oncology
• Bioinformatics
• Genomics

Background:

• Accurate cancer risk stratification is crucial for personalized medicine, enabling targeted prevention and treatment strategies.
• Multi-omics data integration offers a comprehensive approach to identify robust biomarkers for cancer risk assessment.
• Genomic data, including methylation, SCNV, miRNA, and RNAseq, holds significant potential for understanding cancer heterogeneity.

Purpose of the Study:

• To develop and validate a multi-omics framework for stratifying cancer patients into distinct risk groups.
• To identify reliable biomarkers from integrated omics data for predicting patient prognosis and guiding therapeutic decisions.
• To enhance the accuracy of cancer risk assessment through advanced feature extraction and clustering techniques.

Main Methods:

• A novel multi-omics framework utilizing autoencoders for non-linear feature representation and tensor analysis for integrated feature learning.
• Application of clustering methods to stratify patients into multiple cancer risk groups based on extracted latent variables.
• Experimental validation using methylation, somatic copy-number variation (SCNV), micro RNA (miRNA), and RNA sequencing (RNAseq) data from Glioma and Breast Invasive Carcinoma patient cohorts (TCGA dataset).

Main Results:

• The proposed framework successfully extracted informative latent variables from fused multi-omics data, enabling significant patient stratification (p-value<0.05).
• Survival analysis and classification models based on the integrated omics data demonstrated superior performance compared to state-of-the-art methods.
• The study identified distinct cancer risk groups, highlighting the potential for improved clinical decision-making and patient management.

Conclusions:

• The developed multi-omics framework provides a powerful tool for identifying cancer risk groups, facilitating personalized oncology.
• Integrated analysis of diverse omics data significantly improves the accuracy of cancer risk stratification and outcome prediction.
• The open-source availability of this pipeline empowers researchers and clinicians to leverage multi-omics data for enhanced cancer care.