Long-read single-cell genomics: resolving chimeras in multiple displacement amplification

Technical Summary: Long-read Single-Cell Genomics and Chimera Resolution

The Problem
Multiple Displacement Amplification (MDA) is a standard technique for whole-genome amplification (WGA) from single cells, yet it introduces severe technical artifacts that hinder downstream genomic analysis, particularly when paired with long-read sequencing technologies. The primary issue is the generation of chimeric DNA molecules—artificially joined sequences that do not exist in the native genome. These chimeras, combined with highly uneven coverage and coverage dropouts inherent to MDA, lead to extreme rates of misassembly. When standard assembly algorithms designed for bulk sequencing are applied to MDA-derived long-read data, they produce thousands of artifactual structural variants and fragmented, inaccurate genome reconstructions, effectively obscuring the true genomic architecture of single cells.

Methodology
To address these challenges, the authors systematically evaluated the performance of PacBio HiFi long-read sequencing on MDA-amplified DNA from the model green alga Chlamydomonas reinhardtii. Recognizing that conventional assembly tools fail to distinguish between true biological variation and MDA-induced chimeras, the team developed lrSAGA (long-read Single Amplified Genome Assembly). This novel computational tool is specifically engineered to handle the unique error profiles of single-cell long-read datasets. The methodology involved:

1. Benchmarking: Comparing MDA-derived libraries against bulk sequencing standards to quantify chimera rates and coverage bias.
2. Tool Development: Creating lrSAGA to optimize the assembly of long-read MDA data, focusing on resolving chimeric connections and managing coverage gaps.
3. Validation: Testing lrSAGA on diverse datasets, including published Oxford Nanopore and PacBio HiFi data from single or half Caenorhabditis elegans worms, and applying the pipeline to four uncultivated microbial eukaryotes isolated from environmental water samples (an amoeboflagellate from Naegleria, a Bodo flagellate, and two CRuMs supergroup flagellates: Collodictyon triciliatum and Diphylleia rotans).

Key Results
The study reveals that MDA-derived libraries can exhibit chimera rates impacting up to 70% of reads, which, if uncorrected, result in thousands of false structural variants. The application of lrSAGA yielded significant improvements over conventional assembly algorithms:

• Accuracy: Assemblies generated with lrSAGA demonstrated 75–95% fewer misassemblies compared to standard methods.
• Completeness: Despite the limitations imposed by MDA coverage dropouts, the authors successfully assembled up to 68% of the C. reinhardtii genome from a single haploid cell.
• Generalizability: The tool proved effective across different organisms and sequencing platforms (Oxford Nanopore and PacBio HiFi), generating accurate assemblies for C. elegans and high-quality draft genomes for the four uncultivated protists.
• Environmental Application: For the four uncultivated microbial eukaryotes, the approach produced draft genome assemblies estimated to be 70–84% complete, providing the first high-quality genomic data for these specific lineages.

Significance and Claims
The paper posits that the development of lrSAGA is a critical step toward unlocking the genomic diversity of uncultivated microbial eukaryotes. By specifically addressing the chimera artifacts inherent to MDA, the authors demonstrate that long-read single-cell genomics can move beyond model organisms to characterize complex, unculturable environmental microbes. The work claims to provide a robust framework for generating accurate, highly complete genome assemblies from single cells, thereby overcoming the historical barriers of MDA-induced errors that have previously compromised the utility of single-cell long-read sequencing.