Discovering subgroups of patients from DNA copy number data using NMF on compacted matrices

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Identifying patient subgroups with similar genetic profiles is crucial for personalized treatment in complex genetic diseases.
• DNA copy number (CN) variations are key genetic markers in many disorders.
• Standard Non-negative Matrix Factorization (NMF) is computationally intensive for high-dimensional CN data.

Purpose of the Study:

• To develop a data compaction procedure to improve the efficiency and applicability of NMF for analyzing high-dimensional CN data.
• To enable the discovery of biologically relevant patient subgroups from CN profiles.
• To establish robust quality measures for assessing subgroup identification.

Main Methods:

• A novel data compaction technique was developed to reduce the dimensionality of high-resolution CN microarray data.
• Non-negative Matrix Factorization (NMF) was applied to the compacted data for patient subgroup discovery.
• Custom quality measures, inspired by biological relevance, were used alongside standard metrics to evaluate clustering results.

Main Results:

• The proposed procedure significantly improved NMF's applicability to high-dimensional CN data.
• Accurate patient subgroups with distinct molecular and clinical features were identified.
• The method outperformed standard NMF in factorization fitness and identified robust subgroups.

Conclusions:

• The developed data compaction method enhances NMF for discovering patient subgroups based on CN profiles in heterogeneous diseases.
• This approach facilitates more accurate and efficient analysis of complex genetic data for improved treatment strategies.
• The findings support the utility of this method in clinical genomics and personalized medicine.