Data-Driven Identification of Early Cancer-Associated Genes via Penalized Trans-Dimensional Hidden Markov Models

Area of Science:

• Oncology
• Genetics
• Epigenetics

Background:

• Colorectal cancer (CRC) presents a significant global health challenge with high prevalence and mortality.
• Late-stage diagnosis is common, highlighting the need for improved early detection and understanding of CRC progression.
• Epigenetic alterations, particularly DNA methylation, play a crucial role in early cancer stages and gene regulation.

Purpose of the Study:

• To identify differential DNA methylation patterns and key survival-related genes in colorectal cancer.
• To enhance diagnostic accuracy and potentially extend patient survival through novel biomarker discovery.
• To develop and apply advanced statistical methods for analyzing methylation and survival data in CRC.

Main Methods:

• Utilized two novel statistical methods based on sparse, finite mixture of accelerated failure time regression models.
• Analyzed colorectal cancer datasets to identify differentially methylated cytosines and survival-associated genes.
• Compared the performance of the developed methods against two leading existing approaches.

Main Results:

• The developed methods outperformed existing approaches, identifying robust findings and novel differentially methylated sites.
• CRC patient survival was modeled using a two-component mixture regression, indicating distinct patient subgroups.
• Specific genes (CDH11, EPB41L3, DOCK2) were associated with aggressive CRC, while others (TMEM215, PPP1R14A, GPR158, NAPSB) were linked to less aggressive forms.

Conclusions:

• The study successfully identified key genes and DNA methylation patterns associated with colorectal cancer aggressiveness and patient survival.
• The novel analytical methods provide a robust framework for future research in cancer epigenetics and biomarker discovery.
• Findings offer potential for improved diagnostic tools and targeted therapies for colorectal cancer patients.