Integrating Somatic Mutations for Breast Cancer Survival Prediction Using Machine Learning Methods.

Area of Science:

• Computational biology
• Genomics
• Oncology

Background:

• Breast cancer is a leading cause of mortality in women, necessitating improved survival prediction.
• Current predictive models often struggle with accuracy despite using multi-omics data like gene expression.
• The prognostic value of somatic mutations in breast cancer requires further investigation.

Purpose of the Study:

• To develop an accurate breast cancer survival predictive model by integrating somatic mutation data with other molecular data.
• To evaluate the effectiveness of multiple kernel learning (MKL) for multi-omics data integration.
• To assess the contribution of somatic mutations to breast cancer prognosis.

Main Methods:

• Utilized maximum relevance minimum redundancy (mRMR) for feature selection on high-dimensional omics data.
• Employed multiple kernel learning (MKL) to integrate gene expression, copy number variation (CNV), methylation, protein expression, and somatic mutation data.
• Compared the performance of the proposed MKL approach with traditional classifiers.

Main Results:

• The proposed MKL method achieved optimal performance in breast cancer survival prediction.
• Including somatic mutation data led to a remarkable improvement in prediction accuracy.
• mRMR outperformed other feature selection methods, and MKL surpassed traditional classifiers in data integration.

Conclusions:

• Somatic mutations are critical for enhancing breast cancer survival predictions.
• Effective integration of multi-omics data, particularly somatic mutations, using methods like MKL can significantly improve prognostic accuracy.
• This approach offers potential for more optimal clinical diagnosis and treatment strategies for breast cancer patients.