A penalized regression approach for DNA copy number study using the sequencing data.

Area of Science:

• Genomics
• Bioinformatics
• Statistical Modeling

Background:

• High-throughput next-generation sequencing (NGS) data analysis for DNA copy number variant (CNV) detection in tumor and control samples presents significant challenges.
• Accurate profiling of genomic alterations is crucial for understanding cancer biology and developing targeted therapies.

Purpose of the Study:

• To develop an efficient statistical method for detecting multiple CNVs from NGS reads ratio data.
• To introduce a novel Bayesian information criterion (JMIC) for optimal tuning parameter selection in CNV detection models.
• To validate the proposed method and JMIC using simulated and real-world breast cancer sequencing data.

Main Methods:

• A multiple statistical change-points model employing the 1d fused LASSO penalized regression approach for ordered, one-dimensional data.
• A path algorithm to efficiently estimate the number and locations of CNV region boundaries simultaneously.
• Development and comparison of a modified Bayesian information criterion (JMIC) against existing criteria for parameter selection.

Main Results:

• The proposed 1d fused LASSO-based method efficiently detects multiple CNVs from NGS reads ratio data.
• The novel JMIC criterion demonstrates superior performance in tuning parameter selection compared to three established Bayesian information criteria in simulations.
• Application to breast tumor cell line sequencing data yields results consistent with established literature findings.

Conclusions:

• The developed penalized regression approach offers an efficient and accurate method for multiple CNV detection using NGS data.
• The JMIC criterion provides a reliable approach for selecting optimal parameters in CNV analysis, enhancing model performance.
• This methodology contributes to improved genomic profiling for cancer studies, aiding in the identification of clinically relevant CNVs.