Benchmarking pipelines for subclonal deconvolution of bulk tumour sequencing data.

Area of Science:

• Oncology
• Bioinformatics
• Computational Biology

Background:

• Intratumour heterogeneity enables cancer cells to adapt and develop treatment resistance.
• Accurate characterization of tumour clonal architecture is crucial for developing personalized cancer therapies and tracking evolutionary dynamics.
• Existing methods for subclonal deconvolution from bulk sequencing data lack systematic benchmarking, hindering the selection of optimal approaches and understanding dataset impacts.

Purpose of the Study:

• To systematically benchmark subclonal deconvolution pipelines for analysing bulk tumour sequencing data.
• To evaluate the impact of dataset characteristics such as sequencing depth, tumour complexity, and purity on pipeline performance.
• To identify the most accurate subclonal deconvolution pipeline for cancer research.

Main Methods:

• Generation of 80 bulk tumour whole exome sequencing datasets using a comprehensive tumour genome simulation tool.
• Varying dataset characteristics including sequencing depth, tumour complexity, and purity.
• Benchmarking of various subclonal deconvolution pipelines using the simulated datasets.

Main Results:

• Tumour complexity was found to have no significant impact on subclonal deconvolution accuracy.
• Increasing tumour purity and purity-corrected sequencing depth demonstrably improved the accuracy of subclonal deconvolution.
• The optimal subclonal deconvolution pipeline was identified as a combination of Mutect2, FACETS, and PyClone-VI.

Conclusions:

• The study provides a robust benchmark for subclonal deconvolution methods in cancer research.
• Purity and sequencing depth are critical factors influencing the accuracy of tumour clonal architecture analysis.
• The recommended pipeline (Mutect2, FACETS, PyClone-VI) offers a highly accurate approach for subclonal deconvolution, with simulated datasets made publicly available.