MOTIVATION: Accurate demultiplexing of pooled single-cell RNA-seq (scRNA-seq) data is critical for large-scale studies. However, existing methods like vireo, while effective up to ∼16 donors, often struggle with poor clustering due to local optima as donor numbers rise. In high-donor scenarios, overlapping genotypes, a dense genotype space, and increased doublet formation make demultiplexing challenging, requiring methods that are robust to sparse, high-dimensional data and maintain reliable accuracy even as sample complexity grows. RESULTS: We present an enhanced version of souporcell capable of demultiplexing up to 64 donors. The method uses 10× merge for initialization, K-Harmonic Means for robust clustering, and iterative refinement with reinitialization of low-quality clusters and locking of high-quality ones. Compared to vireo, overclustered vireo, and the original souporcell, our approach completely eliminates incorrectly merged clusters and achieves consistently high Adjusted Rand Index (ARI) scores across various doublet rates, demonstrating improved accuracy and scalability. AVAILABILITY AND IMPLEMENTATION: Souporcell3 is freely available under the MIT open-source license at https://github.com/wheaton5/souporcell.

BACKGROUND: Pacific Biosciences (PacBio) circular consensus sequencing (CCS), also known as high fidelity (HiFi) technology, has revolutionized modern genomics by producing long (10 + kb) and highly accurate reads. This is achieved by sequencing circularized DNA molecules multiple times and combining them into a consensus sequence. Currently, the accuracy and quality value estimation provided by HiFi technology are more than sufficient for applications such as genome assembly and germline variant calling. However, there are limitations in the accuracy of the estimated quality scores when it comes to somatic variant calling on single reads. RESULTS: To address the challenge of inaccurate quality scores for somatic variant calling, we introduce TopoQual, a novel tool designed to enhance the accuracy of base quality predictions. TopoQual leverages techniques including partial order alignments (POA), topologically parallel bases, and deep learning algorithms to polish consensus sequences. Our results demonstrate that TopoQual corrects approximately 31.9% of errors in PacBio consensus sequences. Additionally, it validates base qualities up to q59, which corresponds to one error in 0.9 million bases. These improvements will significantly enhance the reliability of somatic variant calling using HiFi data. CONCLUSION: TopoQual represents a significant advancement in genomics by improving the accuracy of base quality predictions for PacBio HiFi sequencing data. By correcting a substantial proportion of errors and achieving high base quality validation, TopoQual enables confident and accurate somatic variant calling. This tool not only addresses a critical limitation of current HiFi technology but also opens new possibilities for precise genomic analysis in various research and clinical applications.

BACKGROUND: Most multiple sequence alignment and string-graph alignment algorithms focus on global alignment, but many applications exist for semi-global and local string-graph alignment. Long reads require enormous amounts of memory and runtime to fill out large dynamic programming tables. Effective algorithms for finding the backbone and thus defining a band of an alignment such as the longest common subsequence with kmer matches (LCSk++) exist but do not work with graphs. This study introduces an adaptation of the Longest Common Subsequence with kmer matches (LCSk++) algorithm tailored for graph structures, particularly focusing on Partial Order Alignment (POA) graphs. POA graphs, which are directed acyclic graphs, represent multiple sequence alignments and effectively capture the relationships between sequences. State-of-the-art methods like ABPOA and SPOA improve upon POA, while ABPOA incorporates banding, SPOA does not; however, neither utilizes parallel processing despite leveraging SIMD for faster matrix calculations. Our approach addresses these limitations by extending the LCSk++ algorithm to handle the complexities of graph-based alignment while incorporating SIMD, banding, and parallel processing for enhanced efficiency. RESULTS: Our extended LCSk++ algorithm integrates dynamic programming and graph traversal techniques to detect conserved regions within POA graphs, termed the LCSk++ backbone. This backbone enables precise banding of the POA matrix for all alignment modes (global, semi-global, and local). Unlike ABPOA, which only allows banded global alignment, our approach enables broader flexibility and significantly improves consensus sequence construction. While supporting more alignment modes than ABPOA, it also outperforms SPOA's global alignment, with substantial memory savings (up to 98%) and significant run-time reductions (up to 25x), particularly for long sequences (> 30,000 bp). Our method maintains high alignment accuracy and proves effective across various string lengths and datasets, including synthetic and PacBio HiFi reads. Parallel processing further enhances runtime efficiency, achieving up to 150x speed improvements on conventional PCs. CONCLUSION: The extended LCSk++ algorithm for graph structures offers a substantial advancement in sequence alignment technology. It effectively reduces memory consumption and optimizes run times without compromising alignment quality, thus providing a robust solution for all alignment modes (global, local, and semi-global) in POA. This method enhances the utility of POA in critical applications such as multiple sequence alignment for phylogeny construction and graph-based reference alignment.

Continual Learning (CL) is an emerging machine learning paradigm in mobile or IoT devices that learns from a continuous stream of tasks. To avoid forgetting of knowledge of the previous tasks, episodic memory (EM) methods exploit a subset of the past samples while learning from new data. Despite the promising results, prior studies are mostly simulation-based and unfortunately do not promise to meet an insatiable demand for both EM capacity and system efficiency in practical system setups. We propose CarM, the first CL framework that meets the demand by a novel hierarchical EM management strategy. CarM has EM on high-speed RAMs for system efficiency and exploits the abundant storage to preserve past experiences and alleviate the forgetting by allowing CL to efficiently migrate samples between memory and storage. Extensive evaluations show that our method significantly outperforms popular CL methods while providing high training efficiency.