CopyMix: Mixture model based single-cell clustering and copy number profiling using variational inference.

Area of Science:

• Genomics
• Computational Biology
• Cancer Research

Background:

• Tumor heterogeneity is crucial for understanding cancer evolution and clinical outcomes.
• Single-cell sequencing technologies enable the identification of distinct tumor cell subpopulations.
• Current computational methods for copy number profiling and clustering are often sequential and prone to artifacts.

Purpose of the Study:

• To develop a novel computational method, CopyMix, for joint inference of cell clusters and copy number profiles from single-cell DNA sequencing data.
• To overcome limitations of sequential analysis and avoid clustering artifacts in tumor subpopulation identification.

Main Methods:

• CopyMix utilizes Variational Inference for a novel mixture model, an advancement of mixture of hidden Markov models.
• The method jointly infers cell clusters and their underlying copy number profiles in a probabilistic graphical model.

Main Results:

• CopyMix demonstrated robust performance on both simulated and biological datasets.
• Evaluation metrics including likelihood-ratio test, CH index, Silhouette, V-measure, and total variation scores support CopyMix's efficacy.

Conclusions:

• CopyMix effectively addresses clustering artifacts by jointly inferring cell clusters and copy number profiles.
• The method shows significant potential for clinical impact in cancer tumor heterogeneity studies.