Computational enhancement of single-cell sequences for inferring tumor evolution

Area of Science:

• Genomics
• Computational Biology
• Cancer Research

Background:

• Single-cell sequencing offers high resolution for assessing single nucleotide variation (SNV) in tumors.
• Current single-cell sequencing technologies generate noisy data with missing bases (MBs) and errors, leading to false positives (FPs) and false negatives (FNs).
• The accuracy of computational methods in correcting these errors and imputing MBs is not well understood.

Purpose of the Study:

• To evaluate the performance of existing and novel computational methods for correcting errors in single-cell tumor sequencing data.
• To assess the accuracy of imputing missing bases and correcting false positives and false negatives.
• To introduce BEAM, a new Bayesian evolution-aware method for improving single-cell sequence quality.

Main Methods:

• Comparative analysis of four existing methods (OncoNEM, SCG, SCITE, SiFit) and a new method (BEAM) using simulated datasets.
• BEAM utilizes molecular phylogenetic frameworks and evolutionary information within single-cell data.
• Evaluation of method performance based on imputation accuracy and correction of FPs/FNs.

Main Results:

• BEAM and SCITE demonstrated the best overall performance in analyzing single-cell tumor sequencing data.
• Most methods accurately imputed missing bases, but effectively detecting and correcting FPs and FNs remains a significant challenge, particularly with smaller datasets.
• Analysis of empirical data confirmed that computational methods can enhance tumor single-cell sequence quality and improve biological inference.

Conclusions:

• Evolutionary continuity in single-cell sequencing datasets, stemming from clonal descent, can be leveraged by methods like BEAM for accurate data imputation.
• While imputation of missing data and base assignments is feasible, correcting false positives and false negatives remains difficult, especially when the number of SNVs is small relative to the number of cells.