O'Connor, S. A. | Search bioRxiv

bioRxiv2024

Breaking out of the cell cycle: Including quiescence in cell cycle classification

Single-cell transcriptomics has unveiled a vast landscape of cellular heterogeneity in which the cell cycle is a significant component. We trained a high-resolution cell cycle classifier (ccAFv2) using single cell RNA-seq (scRNA-seq) characterized human neural stem cells. The ccAFv2 classifies six cell cycle states (G1, Late G1, S, S/G2, G2/M, and M/Early G1) and a quiescent-like G0 state (Neural G0), and it incorporates a tunable parameter to filter out less certain classifications. The ccAFv2 classifier performed better than or equivalent to other state-of-the-art methods even while classifying more cell cycle states, including G0. We demonstrate that the ccAFv2 classifier effectively generalizes the S, S/G2, G2/M, and M/Early G1 states across cell types derived from all three germ layers. While the G0, G1, and Late G1 states perform well in neuroepithelial cell types, their accuracy is lower in other cell types. However, misclassifications are confined to the G0, G1, and Late G1 states. We showcased the versatility of ccAFv2 by successfully applying it to classify cells, nuclei, and spatial transcriptomics data in humans and mice, using various normalization methods and gene identifiers. We provide methods to regress the cell cycle expression patterns out of single cell or nuclei data to uncover underlying biological signals. The classifier can be used either as an R package integrated with Seurat or a PyPI package integrated with SCANPY. We proved that ccAFv2 has enhanced accuracy, flexibility, and adaptability across various experimental conditions, establishing ccAFv2 as a powerful tool for dissecting complex biological systems, unraveling cellular heterogeneity, and deciphering the molecular mechanisms by which proliferation and quiescence affect cellular processes.

cell biology↗

bioRxiv2022

Systematic integration of protein affecting mutations, gene fusions, and copy number alterations into a comprehensive somatic mutational profile

Somatic mutations occur as random genetic changes in genes through protein-affecting mutations (PAMs), gene fusions, or copy number alterations (CNAs). Mutations of different types can have a similar phenotypic effect (i.e., allelic heterogeneity) and should be integrated into a unified gene mutation profile. We developed OncoMerge to fill this niche of integrating somatic mutations to capture allelic heterogeneity, assign a function to mutations, and overcome known obstacles in cancer genetics. Application of OncoMerge to the TCGA Pan-Cancer Atlas increased the number and frequency of somatically mutated genes and improved the prediction of the somatic mutation role as either activating or loss of function. Using integrated somatic mutation matrices increased the power to infer gene regulatory networks and uncovered the enrichment of switch-like feedback motifs and delay-inducing feed-forward loops. These studies demonstrate that OncoMerge efficiently integrates PAMs, fusions, and CNAs and strengthens downstream analyses linking somatic mutations to cancer phenotypes. MotivationGenes are mutated in tumors through either PAMs, CNAs, or gene fusions, thereby splitting the signal of the somatic mutation effect across these three mutation types. We developed OncoMerge to systematically integrate the three mutation types into a single mutation profile that better captures the impact of somatic mutations on cancer phenotypes. As a tool OncoMerge fills the gap between the sophisticated variant calling pipelines and downstream analyses.

systems biology↗