Enhancing metastatic colorectal cancer prediction through advanced feature selection and machine learning techniques

Area of Science:

• Genomics and Bioinformatics
• Oncology
• Machine Learning in Medicine

Background:

• Colorectal cancer (CRC) is a leading global cancer with a high metastasis rate, impacting patient survival.
• Accurate prediction of CRC metastasis is critical for effective treatment and improved outcomes.
• Over 50% of CRC patients develop metastases within five years, necessitating advanced diagnostic tools.

Purpose of the Study:

• To develop and validate a machine learning model for predicting metastatic colorectal cancer (CRC).
• To identify critical genomic biomarkers associated with CRC metastasis using an innovative feature selection algorithm.
• To improve the accuracy of metastasis prediction in CRC, addressing challenges in medical diagnosis.

Main Methods:

• Implementation of a cost-sensitive fast correlation-based filter (CS-FCBF) algorithm for genomic feature selection.
• Reduction of 184 genomic features to 9 critical genes (CXCL9, C2CD4B, RGCC, GFI1, BEX2, CXCL3, FOXQ1, PBK, PLAG1) predictive of metastasis.
• Validation of findings through integrated in vitro, in vivo, and single-cell RNA-seq data analysis.

Main Results:

• The CS-FCBF algorithm significantly enhanced prediction model performance, increasing the area under the precision-recall curve by an average of 21.16%.
• Nine specific genes were identified as crucial predictors of metastatic CRC.
• These identified genes demonstrate potential as diagnostic biomarkers and therapeutic targets for metastatic CRC.

Conclusions:

• Advanced feature selection and machine learning are vital for addressing class imbalance in medical diagnosis, especially for CRC.
• The identified genes are important in the metastatic process of CRC, emphasizing the need for early detection.
• The study's methodology provides valuable insights for future research in oncology and other complex diseases.