BnpC: Bayesian non-parametric clustering of single-cell mutation profiles.

Area of Science:

• Genomics
• Computational Biology
• Cancer Research

Background:

• Single-cell DNA sequencing (scDNA-seq) enables the study of intratumor heterogeneity (ITH).
• Existing methods struggle with large scDNA-seq datasets due to high error rates and missing data, limiting ITH resolution.
• Accurate clonal population identification is crucial for understanding cancer evolution.

Purpose of the Study:

• To introduce BnpC, a novel non-parametric method for clustering cells and inferring genotypes from noisy scDNA-seq data.
• To evaluate BnpC's performance against state-of-the-art methods on simulated and real cancer datasets.
• To provide a scalable and accurate tool for resolving ITH in large scDNA-seq datasets.

Main Methods:

• Development of BnpC, a non-parametric clustering and genotype inference method.
• Comprehensive benchmarking on simulated scDNA-seq data with varying sizes and heterogeneity.
• Application and validation on three real cancer scDNA-seq datasets.

Main Results:

• BnpC demonstrated superior accuracy, runtime, and scalability compared to existing methods on simulated data.
• BnpC achieved the most accurate genotype inference, particularly for highly heterogeneous data.
• BnpC successfully identified previously undetected clonal populations in cancer datasets and handled large datasets (5000 cells).

Conclusions:

• BnpC offers a scalable and accurate solution for analyzing large scDNA-seq datasets.
• The method effectively resolves intratumor heterogeneity by accurately clustering cells and inferring genotypes.
• BnpC is a valuable tool for cancer research and can serve as a preprocessing step to reduce data size.