An Integrated Approach to Knowledge and Prediction Modeling of Breast Cancer Metastasis Using Gene Regulatory...

Area of Science:

• Computational Biology
• Genomics
• Cancer Research

Background:

• Predicting breast cancer metastasis and understanding its mechanisms are critical research goals.
• Current studies often focus on either knowledge discovery or prediction accuracy, rarely integrating both.
• Existing models struggle with accurate prediction and biological explanation of metastasis.

Purpose of the Study:

• To develop a novel computational framework for simultaneous knowledge discovery and explainable prediction of breast cancer metastasis.
• To integrate biological pathway knowledge with machine learning for improved metastasis risk assessment.
• To bridge the gap between understanding metastasis mechanisms and accurate clinical prediction.

Main Methods:

• Construction of gene regulatory networks (GRNs) to model cellular states in metastatic and non-metastatic breast cancer patients.
• Development of a dysregulation score derived from GRN models for explainable metastasis prediction.
• Rigorous evaluation of a model-free classifier leveraging the dysregulation score against complex machine learning models.

Main Results:

• Identification of significant metastasis-associated GRN alterations in breast cancer.
• Discovery of lost co-regulation among key biological processes in metastatic patients.
• Demonstration that the dysregulation score-based classifier outperforms complex machine learning models in prediction accuracy.

Conclusions:

• The proposed framework successfully integrates knowledge discovery and accurate, explainable metastasis prediction.
• The findings reveal critical changes in gene regulatory networks associated with breast cancer metastasis.
• This approach offers potential applicability to other disease contexts requiring similar predictive and explanatory capabilities.